Experimental Study and Modeling of Single - and Two-Phase Flow in Singular Geometries and Safety Relief Valves

Université Libre de Bruxelles
Faculté des sciences appliquées
P h D. D i s s e r t a t i o n
to obtain the title of
Doctorate of Science in Engineering

of the Université Libre de Bruxelles
Specialty: Sciences appliquées
Defended by
Vasilios Kourakos
Experimental study and modeling of

single- and two-phase flow in singular
geometries and safety relief valves
Thesis Supervisors:
Jean-Marie Buchlin
Patrick Rambaud
Saïd Chabane
prepared at Bruxelles, Belgium

defended on October 28, 2011
Jury: Jean-Marie Buchlin, Patrick Rambaud, Saïd Chabane, Jürgen

Schmidt, Pierre Colinet, Benoît Haut, Pascal Souquet
Promotor: Jean-Marie Buchlin - VKI & ULB
President: Pierre Colinet - ULB
Secretary: Benoît Haut - ULB
Reviewers: Jean-Marie Buchlin - VKI & ULB
Patrick Rambaud - VKI
Saïd Chabane - CETIM
Pierre Colinet - ULB
Benoît Haut - ULB
Invited: Jürgen Schmidt - KIT
Pascal Souquet - CETIM
Contact information:
Vasilios Kourakos
von Karman Institute for Fluid Dynamics
Chaussée de Waterloo 72
B-1640, Rhode-Saint-Genèse
BELGIUM
email: kourakos@vki.ac.be
Dedication
To my familly
"A man goes to knowledge as he goes to war:

wide-awake, with fear, with respect, and with absolute
assurance. Going to knowledge or going to war in
any other manner is a mistake, and whoever makes
it might never live to regret it."
Carlos Castaneda (1925–1998),
Peruvian anthropologist and author
Acknowledgements
It has been a very long path for me to reach this point and, as in every long, painful
but fruitful step in life, the help and support from colleagues, friends and family was
indispensable to accomplish this PhD thesis.
First of all, I would like to thank the von Karman Institute (VKI) which has been
the place that offered me the chance to carry out such an interesting research project
and learn so many things. The ambiance of VKI will always stay in my mind with the
best memories. I am extremely grateful as well to the Centre Technique des Industries
Mécaniques (CETIM) for financing the whole project and integrating me in their team
providing at the same time a very friendly working environment.
I am deeply indebted to the three persons that have always supported me and helped
me with their technical and personal advices: Dr. Saïd Chabane from CETIM and Pro-
fessors Patrick Rambaud and Jean-Marie Buchlin from VKI. I am also really gratified for
the great moments we have shared out of working hours during the numerous business
trips which helped me to relax from the stress of work and feel much more comfortable.
I would like to express my gratitude to Pascal Souquet, Pascal François, Muriel Maque-
nnehan and Daniel Pierrat from CETIM for their great support. Furthermore, I want to
acknowledge Professor Jürgen Schmidt from Karlsruhe Institute of Technology (KIT) and
Professor Pierre Colinet and Benoît Haut from Université Libre de Bruxelles (ULB) for
accepting to become members of my PhD jury and for reading in detail my thesis and
providing constructive advices for its improvement.
A word of thanks to the technical team of VKI (Guillaume Diquas, Didier Welter) and
of CETIM (Alain Pilorge, Luc Laundry and François Corbin); the latter has to be par-
ticularly acknowledged since he performed an important part of the SRV measurements.
Both teams have put great effort on the experimental part of the thesis. Numerous students
have been working in the frame of this project and their contribution should be certainly
recognized hereunder: Vaggelis Bacharoudis, Rosario Delgado-Tardáguila, Emrah Deniz,
Bugra Kilinç, Vitor Fernandez. Likewise, the help of Remi Berger (in preliminary PIV
tests) and Flora Tomasoni (for probe processing) should be mentioned.
A special consideration should be given to my friends from VKI for showing me every
moment their true friendship: my real friend Memo (Mehmet Mersilingil), Fabio Pinna
and Delphine Laboureur. The time in VKI has become much more pleasant due to their
presence. The long discussions with my VKI colleagues helping to relax from work have
been so helpful for me, hence I want to thank Benoît, Kostas, Boris, Jan (also for being
a so quite office mate) and the other very good friends I made in VKI. A big word of
thanks to all my friends from Brussels but mainly; Adelaida, Margaritis, Dimitris, Merve,
Yannis, Vaggelis for making me love and miss so much such a rainy country as Belgium.
I should not forget to mention my true friends from Greece who, although I don’t
see them so often any more, never stopped caring for me and have always been close to
me; Giwrgos, Eftuxia, Vaggelis and Dimitris (regardless of his long absence); I hope that
soon I will be back home close to them. I want also to thank so much Katerina for her
support and patience and for being always there for me in this difficult period for both of
us. Moreover, a word of thanks to my professor from Greece Professor D. P. Margaris
from University of Patras who, few years ago, has triggered my interest for the wonderful
world of fluid dynamics.
Je voudrais specialement dédier ce paragraph à Saïd Chabane et François Corbin. Je
serais toujours ravis pour l’intimité qu’ils m’ont offert aussi généreusement. Je garde et
je ne garderai que des excellents souvenirs de leur accueil à Nantes et au CETIM. Je suis
sûr que notre amitié restera impacte.
Finally, most of all I want to thank, with a few words in Greek, my family who has
always been my major inspiration in life;
θα ήθελα, τέλος, να ευχαριστήσω την οικογένειά μου που με έχει στηρίξει

τόσο πολύ σε κάθε μου απόφαση μέχρι τώρα· με κάθε τρόπο και με έμπρακτη
και ειλικρινή αγάπη...
Abstract
This research project was carried out at the von Karman Institute for Fluid Dynamics
(VKI), in Belgium, in collaboration and with the funding of Centre Technique des Indus-
tries Mécaniques (CETIM) in France.
The flow of a mixture of two fluids in pipes can be frequently encountered in nuclear,
chemical or mechanical engineering, where gas-liquid reactors, boilers, condensers, evap-
orators and combustion systems can be used. The presence of section changes or more
generally geometrical singularities in pipes may affect significantly the behavior of two-
phase flow and subsequently the resulting pressure drop and mass flow rate. Therefore, it
is an important subject of investigation in particular when the application concerns indus-
trial safety valves.
This thesis is intended to provide a thorough research on two-phase (air-water) flow
phenomena under various circumstances. The project is split in the following steps. At
first, experiments are carried out in simple geometries such as smooth and sudden diver-
gence and convergence singularities. Two experimental facilities are built; one in smaller
scale in von Karman Institute and one in larger scale in CETIM. During the first part of
the study, relatively simple geometrical discontinuities are investigated. The characteri-
zation and modeling of contraction and expansion nozzles (sudden and smooth change of
section) is carried out. The pressure evolution is measured and pressure drop correlations
are deduced. Flow visualization is also performed with a high-speed camera; the different
flow patterns are identified and flow regime maps are established for a specific configura-
tion. A dual optical probe is used to determine the void fraction, bubble size and velocity
upstream and downstream the singularities.
In the second part of the project, a more complex device, i.e. a Safety Relief Valve
(SRV), mainly used in nuclear and chemistry industry, is thoroughly studied. A transpar-
ent model of a specific type of safety valve (1 1/2" G 3" ) is built and investigated in terms
of pressure evolution. Additionally, flow rate measurements for several volumetric quali-
ties and valve openings are carried out for air, water and two-phase mixtures. Full optical
access allowed identification of the structure of the flow. The results are compared with
measurements performed at the original industrial valve. Flowforce analysis is performed
revealing that compressible and incompressible flowforces in SRV are inversed above a
certain value of valve lift. This value varies with critical pressure ratio, therefore is di-
rectly linked to the position at which chocked flow occurs during air valve operation. In
two-phase flow, for volumetric quality of air β=20%, pure compressible flow behavior, in
terms of flowforce, is remarked at full lift. Numerical simulations with commercial CFD
code are carried out for air and water in axisymmetric 2D model of the valve in order to
verify experimental findings.
The subject of modeling the discharge through a throttling device in two-phase flow
is an important industrial problem. The proper design and sizing of this apparatus is a
crucial issue which would prevent its wrong function or accidental operation failure that
could cause a hazardous situation. So far reliability of existing models predicting the
pressure drop and flow discharge in two-phase flow through the valve for various flow
conditions is questionable. Nowadays, a common practice is widely adopted (standard
ISO 4126-10 (2010), API RP 520 (2000)); the Homogeneous Equilibrium Method with
the so-called ω-method, although it still needs further validation. Additionally, based on
ω-methodology, Homogeneous Non-Equilibrium model has been proposed by Diener and
Schmidt (2004) (HNE-DS), introducing a boiling delay coefficient. The accuracy of the
aforementioned models is checked against experimental data both for transparent model
and industrial SRV. The HNE-DS methodology is proved to be the most precise among the
others. Finally, after application of HNE-DS method for air-water flow with cavitation,
it is concluded that the behavior of flashing liquid is simulated in such case. Hence, for
the specific tested conditions, this type of flow can be modeled with modified method
of Diener and Schmidt (CF-HNE-DS) although further validation of this observation is
required.
Contents
List of figures vii
List of tables xv
List of symbols xvii
Abbreviations xxi
I Preface 1
1 Fundamentals of the study 3
1.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.1.1 Basics of two-phase flows . . . . . . . . . . . . . . . . . . . . . 3
1.1.2 Two-phase flow occurring in industrial devices and safety valves . 4
1.2 Objectives-Achievements / Motivation . . . . . . . . . . . . . . . . . . . 7
1.3 Methodology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
1.4 Overview of the thesis . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
2 Theoretical background-Literature review 15

2.1 Two-phase flows generalities . . . . . . . . . . . . . . . . . . . . . . . . 15
2.1.1 Parameters describing gas-liquid flows . . . . . . . . . . . . . . 15
2.1.2 Description of flow regimes and flow regime maps . . . . . . . . 18
2.1.2.1 Horizontal flow regimes . . . . . . . . . . . . . . . . . 19
2.1.2.2 Flow pattern charts . . . . . . . . . . . . . . . . . . . 20
2.2 Pressure head loss and modeling in straight pipes . . . . . . . . . . . . . 22
2.2.1 Pressure drop and modeling in two-phase flow . . . . . . . . . . 23
2.2.1.1 Different types of two-phase flow models . . . . . . . . 24
2.2.1.2 Homogeneous flow model . . . . . . . . . . . . . . . . 24
2.2.1.3 Separated flow models . . . . . . . . . . . . . . . . . 24
2.3 Flow in geometrical singularities . . . . . . . . . . . . . . . . . . . . . . 26
2.3.1 Description of different types of singularities . . . . . . . . . . . 26
2.3.2 Modeling of pressure change in singularities . . . . . . . . . . . 27
2.3.2.1 Sudden enlargement . . . . . . . . . . . . . . . . . . . 27
2.3.2.2 Sudden contraction . . . . . . . . . . . . . . . . . . . 30
2.3.2.3 Smooth change of cross-section . . . . . . . . . . . . . 30
2.3.3 Literature survey on geometrical singularities . . . . . . . . . . . 31
ii Table of contents
2.3.3.1 Expansion geometries . . . . . . . . . . . . . . . . . . 31

2.3.3.2 Contraction geometries . . . . . . . . . . . . . . . . . 32
2.4 Industrial valves . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
2.4.1 Description of different types of valves . . . . . . . . . . . . . . 32
2.4.2 Introduction to Pressure Relief Valves (PRV) . . . . . . . . . . . 33
2.4.3 Relief Devices (RD)-detailed analysis . . . . . . . . . . . . . . . 35
2.4.3.1 Safety Relief Valve (SRV) important parameters . . . . 35
2.4.3.2 Pilot Operated Relief Valve (PORV) . . . . . . . . . . 38
2.4.3.3 Balanced Bellows Valve . . . . . . . . . . . . . . . . . 38
2.4.4 Single-phase flow aspect of SRV . . . . . . . . . . . . . . . . . 39
2.4.4.1 Incompressible fluid-Liquid . . . . . . . . . . . . . . 40
2.4.4.2 Compressible fluid-Gas . . . . . . . . . . . . . . . . . 40
2.4.5 Cavitation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
2.4.6 Critical flows . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
2.4.7 Existing methodology for two-phase flow in SRV . . . . . . . . 44
2.4.7.1 Homogeneous Equilibrium Model . . . . . . . . . . . 45
2.4.7.2 The Omega method . . . . . . . . . . . . . . . . . . . 48
2.4.7.3 HNE-DS method . . . . . . . . . . . . . . . . . . . . 50
2.4.7.4 Modified CF-HNE-DS method . . . . . . . . . . . . . 52
2.4.7.5 API 520 method . . . . . . . . . . . . . . . . . . . . . 53
2.4.7.6 Comparison of models . . . . . . . . . . . . . . . . . 53
2.4.8 Literature review on SRV . . . . . . . . . . . . . . . . . . . . . 55
2.4.8.1 Experimental studies-Modeling . . . . . . . . . . . . . 55
2.4.8.2 Numerical studies in single-phase flow . . . . . . . . . 58
2.5 Conclusions for Chapter 2 . . . . . . . . . . . . . . . . . . . . . . . . . 60
II Measurement campaign 63
3 Experimental techniques-facilities 65
3.1 Characterization of facilities for singularities . . . . . . . . . . . . . . . 65
3.1.1 Small scale . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
3.1.1.1 Pressure measurements . . . . . . . . . . . . . . . . . 68
3.1.1.2 Optical probe-visualization . . . . . . . . . . . . . . . 68
3.1.1.3 Test conditions small scale . . . . . . . . . . . . . . . 68
3.1.2 Large scale . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70
3.1.2.1 Pressure . . . . . . . . . . . . . . . . . . . . . . . . . 70
3.1.2.2 Test conditions large scale . . . . . . . . . . . . . . . 72
3.1.3 Summarizing tests and flow conditions . . . . . . . . . . . . . . 73
3.2 Characterization of facilities for SRV . . . . . . . . . . . . . . . . . . . 73
3.2.1 SRV pressure study . . . . . . . . . . . . . . . . . . . . . . . . . 75
3.2.1.1 SRV transparent facility . . . . . . . . . . . . . . . . . 75
3.2.1.2 Industrial SRV facility . . . . . . . . . . . . . . . . . . 78
3.2.1.3 Test conditions SRV . . . . . . . . . . . . . . . . . . 82
3.2.2 SRV optical probe and visualization study (LUCY III) . . . . . . 83
3.3 Measurement techniques . . . . . . . . . . . . . . . . . . . . . . . . . . 84
ii
Table of contents iii
3.3.1 Flow meters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85

3.3.2 Pressure measurements . . . . . . . . . . . . . . . . . . . . . . . 86
3.3.3 Optical probe . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
3.3.3.1 Void fraction . . . . . . . . . . . . . . . . . . . . . . . 87
3.3.3.2 Bubble velocity and bubble diameter . . . . . . . . . . 88
III Geometrical singularities 93

4 Expansion singularities 95
4.1 Pressure measurements . . . . . . . . . . . . . . . . . . . . . . . . . . . 95
4.1.1 Determination of pressure drop coefficient . . . . . . . . . . . . . 96
4.1.2 Sudden expansion . . . . . . . . . . . . . . . . . . . . . . . . . 97
4.1.3 Comparison smooth-sudden expansion . . . . . . . . . . . . . . 99
4.1.4 Summary of expansion pressure measurements . . . . . . . . . . 102
4.1.4.1 3D comparative plots in expansion geometries . . . . . 104
4.1.4.2 Developing length in expansion geometry . . . . . . . 106
4.2 Flow visualization in expansion singularities . . . . . . . . . . . . . . . 108
4.3 Optical probe measurements in divergent section . . . . . . . . . . . . . 111
4.4 Comparison with existing models . . . . . . . . . . . . . . . . . . . . . 120
4.5 New correlations proposed . . . . . . . . . . . . . . . . . . . . . . . . . 122
5 Contraction singularities 127

5.1 Pressure results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 127
5.1.1 Sudden-smooth contraction . . . . . . . . . . . . . . . . . . . . 127
5.1.2 Summary of contraction pressure results . . . . . . . . . . . . . . 132
5.2 Visualization in contraction singularities . . . . . . . . . . . . . . . . . . 134
5.3 Comparison with literature models . . . . . . . . . . . . . . . . . . . . . 135
5.4 Application of Omega-method in contraction geometry . . . . . . . . . . 136
5.5 New correlation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 137
IV Safety Relief Valve 141

6 Safety Relief Valve 143
6.1 Pressure and flow rate measurements . . . . . . . . . . . . . . . . . . . 143
6.1.1 Industrial valve-transparent model discrepancies . . . . . . . . . 143
6.1.2 Single and two-phase air-water flow . . . . . . . . . . . . . . . . 144
6.2 Flow visualization in SRV . . . . . . . . . . . . . . . . . . . . . . . . . 144
6.3 Optical probe measurements in SRV . . . . . . . . . . . . . . . . . . . . 147
6.4 Flowforce in SRV . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 151
6.4.1 Axisymmetric theoretical analysis of flowforce in SRV . . . . . . 152
6.4.2 Experimental investigation . . . . . . . . . . . . . . . . . . . . . 155
6.4.3 CFD simulations . . . . . . . . . . . . . . . . . . . . . . . . . . 160
iii
iv Table of contents
6.4.3.1 CFD results . . . . . . . . . . . . . . . . . . . . . . . 161

6.4.3.2 Comparison flowforce experiments-CFD . . . . . . . . 164
6.4.3.3 Influence of adjustment ring position . . . . . . . . . . 166
6.5 Pressure drop in SRV . . . . . . . . . . . . . . . . . . . . . . . . . . . . 167
6.6 Mass flux through SRV-discharge coefficient . . . . . . . . . . . . . . . . 168
6.6.1 Non-flashing-“Frozen” two-phase flow . . . . . . . . . . . . . . 171
6.6.2 Flashing two-phase flow . . . . . . . . . . . . . . . . . . . . . . 174
6.7 Correlation for pressure drop in SRV . . . . . . . . . . . . . . . . . . . 177
V Conclusions and future recommendations 181
7 Discussion 183
7.1 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 183
7.2 Future works-recommendations . . . . . . . . . . . . . . . . . . . . . . 186
Bibliography 187
Appendices 197
Appendix A Uncertainty analysis 199

A.1 Pressure drop coefficient formula . . . . . . . . . . . . . . . . . . . . . . 199
Appendix B Design of experimental facilities 203
Appendix C Disasters caused by PRV failures 209
Appendix D PED Annex II diagrams 213
Appendix E PRV orifice sizes 217
Appendix F Vapor-liquid flow 219
Appendix G Additional flow regime charts 221

G.1 Vertical upward cocurrent flow . . . . . . . . . . . . . . . . . . . . . . . 221
G.1.1 Flow regimes . . . . . . . . . . . . . . . . . . . . . . . . . . . . 221
G.1.2 Flow maps . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 222
G.2 Vertical downward cocurrent flow . . . . . . . . . . . . . . . . . . . . . 223
G.2.1 Flow patterns . . . . . . . . . . . . . . . . . . . . . . . . . . . . 224
G.2.2 Flow regime maps . . . . . . . . . . . . . . . . . . . . . . . . . 225
G.3 Slightly inclined pipe flow . . . . . . . . . . . . . . . . . . . . . . . . . 225
Appendix H Pressure drop in straight pipe SP flow 229
iv
Table of contents v
Appendix I Straight pipe two-phase flow models 233

I.1 Martinelli-Nelson [1948] model . . . . . . . . . . . . . . . . . . . . . . 233
I.2 Drift-flux model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 235
I.3 Thom [1964] model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 236
I.4 Baroczy model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 237
I.5 Chisholm [1976] model . . . . . . . . . . . . . . . . . . . . . . . . . . . 238
I.6 Evaluation of the models . . . . . . . . . . . . . . . . . . . . . . . . . . 239
I.7 Flow pattern dependent models . . . . . . . . . . . . . . . . . . . . . . . 240
Appendix J Single-phase pressure change in singularities 243

J.1 Abrupt area changes . . . . . . . . . . . . . . . . . . . . . . . . . . . . 244
J.2 Calculations in geometries with change of section . . . . . . . . . . . . . 244
J.2.1 Single-phase pressure drop . . . . . . . . . . . . . . . . . . . . 246
J.2.2 Two-phase pressure drop . . . . . . . . . . . . . . . . . . . . . 247
J.2.3 Pressure drop coefficient in contractions . . . . . . . . . . . . . 247
Appendix K Pressure drop database 249

K.1 Expansion singularities . . . . . . . . . . . . . . . . . . . . . . . . . . . 249
K.2 Contraction singularities . . . . . . . . . . . . . . . . . . . . . . . . . . 257
K.3 SRV pressure drop database . . . . . . . . . . . . . . . . . . . . . . . . 261
Appendix L Matlab codes 265

L.1 Singularities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 265
L.2 SRV . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 270
v
List of Figures
1.1 Different types of pipe fittings-geometrical singularities (SS Engineers &

Consultants (2010)). . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.2 Detailed sketch of nuclear reactor core. . . . . . . . . . . . . . . . . . . 6
1.3 Schematic of the steam system and components typical of a large turbine
unit (Taken by Bereznai (2005)). . . . . . . . . . . . . . . . . . . . . . . 6
1.4 Safety Relief Valves tested under cryogenic conditions in Kennedy Space
Center. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
1.5 Cutaway of PRV. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
1.6 Schematic of the problem. . . . . . . . . . . . . . . . . . . . . . . . . . 9
1.7 Deepwater Horizon oil spill accident (2010). . . . . . . . . . . . . . . . . 9
1.8 Methodology steps for geometrical singularities study. . . . . . . . . . . 11
1.9 Plan of transparent modified SRV (1 1/2 G 3 ). . . . . . . . . . . . . . .
00 00
12
1.10 Methodology steps for Safety Relief Valve study. . . . . . . . . . . . . . 13
2.1 Flow patterns in horizontal flow in a pipe. Adapted by Weisman (1983). . 21

2.2 Images of flow regimes in horizontal flow in a pipe (taken from Ghajar
(2004)). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
2.3 Horizontal flow map of Baker modified by Bell et al. (1970). . . . . . . . 22
2.4 Two-phase flow multiplier Φ2L function of the Lockhart and Martinelli
(1949) parameter. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
2.5 Different types of singular geometries. . . . . . . . . . . . . . . . . . . . 28
2.6 Sudden contraction geometry . . . . . . . . . . . . . . . . . . . . . . . . 30
2.7 Different types of valves. . . . . . . . . . . . . . . . . . . . . . . . . . . 33
2.8 Conventional spring loaded SRV. . . . . . . . . . . . . . . . . . . . . . . 34
2.9 Pilot operated valve (PORV). . . . . . . . . . . . . . . . . . . . . . . . 38
2.10 Balanced bellows (left) and balanced piston RV (right) Tyco (2008). . . . 39
2.11 Critical point diagram of water. Cavitation and boiling phenomena. . . . . 42
2.12 Examples of cavitation. . . . . . . . . . . . . . . . . . . . . . . . . . . 43
2.13 Speed of sound in two-phase flow medium. . . . . . . . . . . . . . . . . 44
2.14 Nozzle flow. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
2.15 Comparison of discharged relief area calculated with different two-phase
flow models (from Tran and Reynolds (2007)). . . . . . . . . . . . . . . 55
2.16 Flowforce characteristic of a PRV taken from Föllmer and Schnettler
(2003). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
2.17 Results obtained by Song et al. (2010) . . . . . . . . . . . . . . . . . . . 59
2.18 Velocity iso-surfaces for water flow in PRV taken from Vallet et al. (2010). 60
viii List of figures
3.1 A) Progressive expansion of different opening angles-reattachment length

L/d.
B) Sudden expansion-reattachment length L/d. . . . . . . . . . . . . . . . 67
3.2 Experimental small scale LUCY II facility. . . . . . . . . . . . . . . . . 67
3.3 Gas injectors used in small scale setup. . . . . . . . . . . . . . . . . . . 68
3.4 Convergence test section for pressure measurements in small scale. . . . 69
3.5 Test section for optical probe measurements in small scale. . . . . . . . . 69
3.6 Gas injectors used in large scale facility. . . . . . . . . . . . . . . . . . . 71
3.7 Experimental large scale AGATHE facility. . . . . . . . . . . . . . . . . 72
3.8 Test section (AGATHE). . . . . . . . . . . . . . . . . . . . . . . . . . . 73
3.9 Experimental setup to study SRV model (AGATHE II). . . . . . . . . . . 76
3.10 Gas injector DN40 (with more holes). . . . . . . . . . . . . . . . . . . . 76
3.11 Schematic of the test section AGATHE II (left) and comparison of indus-
trial safety valve with the transparent model (right). . . . . . . . . . . . . 77
3.12 Picture of the test section AGATHE II and its different components. . . . 77
3.13 Transparent valve body (left) valve disk with three pressure taps (right). . 78
3.14 Transparent SRV and position of pressure taps. . . . . . . . . . . . . . . 79
3.15 Experimental setup for SRV water study (taken from Corbin et al. (2009a)).
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
3.16 Experimental installation for compressible fluid SRV study (taken from
Chabane et al. (2009)). . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
3.17 Modified industrial water SRV facility for flowforce measurements in
two-phase flow. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81
3.18 Gas injector DN80 for industrial SRV facility. . . . . . . . . . . . . . . . 81
3.19 Relative position of valve seat and adjustment ring. . . . . . . . . . . . . 82
3.20 Transparent SRV setup for optical probe measurements and flow visual-
ization (LUCY III). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84
3.21 Test section of LUCY III with detailed view of the optical probe position-
ing. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84
3.22 Working principle of electromagnetic flow meter (from Copa-XE DE43F).
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86
3.23 Variable reluctance differential pressure transducer(Validyne (2005)). . . 87
3.24 Working principle of optical probe. . . . . . . . . . . . . . . . . . . . . 88
3.25 Probe tips inside flow. . . . . . . . . . . . . . . . . . . . . . . . . . . . 89
4.1 Explanation of the way to determine the singular pressure change in ex-
pansion geometry. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96
4.2 Single-phase static pressure change versus axial position for sudden en-
largement of σ=0.43 and for ReL1 =8.4 ·105 . Comparison of experimental
results with Idel’Cik (1986) calculation. . . . . . . . . . . . . . . . . . . 98
4.3 Two-phase static pressure change versus axial position for sudden en-
largement of σ=0.43 and for ReL1 =1.82·105 -comparison with experimen-
tal single-phase and with models of Jannsen and Kervinen (1966) and
Chisholm (1969). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98
viii
List of figures ix
4.4 Dimensionless singular pressure change ΦL versus volumetric quality.

Comparison with models of Jannsen and Kervinen (1966) and Chisholm
(1969). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99
4.5 ∆P sing for several ReL1 from 0-18% of air for sudden enlargement of sur-
face areas σ=0.43 and σ=0.65. . . . . . . . . . . . . . . . . . . . . . . . 100
4.6 Comparison of pressure distribution for different expansion geometries in
single and two-phase flow. . . . . . . . . . . . . . . . . . . . . . . . . . 101
4.7 Pressure evolution along expansion geometries; the position of recircula-
tion eddy can be identified. . . . . . . . . . . . . . . . . . . . . . . . . . 102
4.8 Pressure drop coefficient ζ versus volumetric quality for different ReL1 in
sudden expansion σ=0.43 . . . . . . . . . . . . . . . . . . . . . . . . . . 103
smooth expansion σ=0.43, α=15 ◦ . . . . . . . . . . . . . . . . . . . . . . 103
4.12 Two-phase multiplier ΦL versus β for various expansion geometries. . . . 105
4.13 3D comparison plot of expansion singularities. . . . . . . . . . . . . . . 106
4.14 Effect of opening expansion angle on the pressure drop coefficient for
various upstream mass fluxes. . . . . . . . . . . . . . . . . . . . . . . . 107
4.15 Pressure drop coefficient for different opening angles of a diffuser (left),
flow inside diffuser (right)-taken from Comolet (1963). . . . . . . . . . . 107
4.16 Developing length versus opening angle for enlargement singularity. . . . 108
4.17 Experimentally determined developing length in sudden expansion for
different σ-comparison with Ahmed et al. (2008) correlation . . . . . . . 108
4.18 Developing length in expansion singularity versus surface area ratio pro-
posed by different authors (taken from Ahmed et al. (2008)) . . . . . . . 108
4.19 Flow patterns identified downstream of the divergence geometry of α=9 ◦
and σ=0.64. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109
4.20 Visualization in enlargement geometries for ReL1 =1.85·105 and 7% volu-
metric quality of air. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 110
4.21 Modified Baker (1954) map for progressive and sudden expansion of
σ=0.43 and 0.65. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111
4.22 Flow map for progressive expansion of σ=0.64 and α=9 ◦ . . . . . . . . . 112
4.23 Horizontal and vertical profiles upstream and downstream divergence sec-
tion. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113
4.24 Flow visualization upstream and downstream divergence section-comparison
with void fraction profiles. . . . . . . . . . . . . . . . . . . . . . . . . . 114
4.25 Example of bubble size distribution diagram with log-normal fit, peak
diameter D peak identified. . . . . . . . . . . . . . . . . . . . . . . . . . . 115
4.26 Bubble size distribution at the horizontal plane for β=6% and 9% at up-
stream and downstream positions for different locations in the pipe. . . . 116
4.27 Influence of volumetric quality on bubble size distribution at the center of
the pipe upstream and downstream the singularity. . . . . . . . . . . . . . 117
ix
x List of figures
4.28 Comparison of void fraction profiles obtained for two different gas injec-
tors (taken from Deniz et al. (2009)). . . . . . . . . . . . . . . . . . . . 118
4.29 Bubble and liquid movement in the pipe. . . . . . . . . . . . . . . . . . . 118
4.30 Comparison of horizontal and vertical bubble velocity profiles upstream
and downstream the singularity for β=9% and 14%. . . . . . . . . . . . 119
4.31 Explanation of blockage effect in upper part of the duct and bubble veloc-
ity versus chord length diagram. . . . . . . . . . . . . . . . . . . . . . . 119
4.32 Deviation of Jannsen and Kervinen (1966) model from experimental results.121
4.33 Deviation of Chisholm (1969) model from experimental results. . . . . . 121
4.34 Deviation of predicted-measured pressure drop coefficient for expansion
singularities of σ=0.43. . . . . . . . . . . . . . . . . . . . . . . . . . . 123
4.35 Deviation of predicted-measured pressure drop coefficient for expansion
singularities of σ=0.65. . . . . . . . . . . . . . . . . . . . . . . . . . . . 123
4.36 Correlation for all expansion geometries. . . . . . . . . . . . . . . . . . . 124
5.1 Explanation of the way to determine the singular pressure change in con-
traction geometry. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 128
5.2 Experimental and numerical single and two-phase static pressure change
versus axial position for convergence of σ=1.56 and 9 ◦ angle for several
ReL1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 129
5.3 Single-phase static ∆P sing obtained experimentally and numerically for
several Q̄water in progressive convergence geometry of σ=1.56 and 9 ◦ angle.130
5.4 Experimental and numerical dimensionless singular static pressure change
ΦLst versus volumetric quality. Comparison to literature (Guglielmini et al.
(1997)) and to adapted (C=0.81) Jannsen and Kervinen (1966) model. . . 130
5.5 Comparison of experimental single-phase water results with Idel’Cik (1986)
calculation for sudden and smooth contraction of σ=2.34, 15 ◦ angle and
ReL1 =2.2·105 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 131
5.6 Two-phase flow results in for sudden and smooth contraction of σ=2.34
and 15 ◦ angle for 9% of air and ReL1 =2.2·105 . Comparison with experi-
mental single-phase and Comolet (1963) formula. . . . . . . . . . . . . . 132
5.7 Pressure drop coefficient in contraction singularities. . . . . . . . . . . . 134
5.8 Two-phase multiplier ΦL for various ReL1 in sudden and smooth contraction.135
5.9 3D plots of ReL1 -β-ζ for sudden contraction σ=2.34, smooth contraction
σ=2.34 and α=15 ◦ and smooth contraction σ=1.56 and α=9 ◦ . . . . . . 136
5.10 Visualization in contraction geometries for ReL1 =1.85·105 and 7% volu-
metric quality of air. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 137
5.11 Predicted with ω-method versus experimental pressure drop coefficient
against mass quality in sudden contraction of σ=2.34 . . . . . . . . . . . 138
5.12 Correlation for contraction σ=2.34. . . . . . . . . . . . . . . . . . . . . 139
5.13 Correlation for smooth contraction of σ=1.56 and 9 ◦ angle. . . . . . . . 140
5.14 Comparison of prediction of ζ with overall correlation for contraction
geometries with experimental measurements. . . . . . . . . . . . . . . . 140
6.1 Volume flow rate of water versus valve opening for transparent and indus-
trial valve. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 144
x
List of figures xi
6.2 Mass flow rate versus β and L in industrial valve at P set =0.3 MPa. . . . . 145
6.3 Cavitation in SRV for P set =0.3 MPa at valve openings L=2, 4.5 and 7.3
mm. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 145
6.4 Flow visualization in the safety valve model-upstream (left) and down-
stream of the valve (right). . . . . . . . . . . . . . . . . . . . . . . . . . 146
6.5 Flow visualization in the core of the safety valve. . . . . . . . . . . . . . 147
6.6 Possible location of cavitation appearance predicted by CFD for various
lifts at P set =0.3 MPa . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 147
6.7 Direction of upstream-downstream SRV optical probe profiles. . . . . . . 148
6.8 Upstream horizontal void fraction profile in SRV for β=2%. . . . . . . . 149
6.9 Upstream vertical void fraction profile in SRV for β=2%. . . . . . . . . . 149
6.10 Upstream horizontal void fraction profile in SRV for β=9%. . . . . . . . 150
6.11 Upstream vertical void fraction profile in SRV for β=9%. . . . . . . . . . 151
6.12 Downstream vertical void fraction profile in SRV for β=2%. . . . . . . . 151
6.13 Downstream vertical void fraction profile in SRV for β=9%. . . . . . . . 152
6.14 Simplified schematic of valve disk with nozzle. . . . . . . . . . . . . . . 153
6.15 Comparison of experimental and theoretical hydrodynamic force of SRV
versus valve opening for P1 =0.3 MPa in water flow. . . . . . . . . . . . . 154
6.16 Comparison of experimental and theoretical hydrodynamic force of SRV
versus valve opening P1 =0.3 MPa in two-phase flow conditions (β=20%). 155
6.17 Force applied on the valve disk versus valve opening for transparent and
industrial valve. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 156
6.18 Influence of adjustment ring position on flowforce for P set =1.21 MPa. . . 156
6.19 3D Flowforce characteristic (P, L, F) plots in industrial SRV. . . . . . . . 157
6.20 Force applied on the valve disk versus valve opening for water, air and
various air-water mixtures . . . . . . . . . . . . . . . . . . . . . . . . . 159
6.21 Relative discrepancy between flowforce of air and two-phase and single-
phase water at P set =0.3 MPa. . . . . . . . . . . . . . . . . . . . . . . . . 160
6.22 Comparison of inverse flowforce position for different set pressures in
industrial SRV. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 161
6.23 Axisymmetric grid of industrial SRV for CFD computations. . . . . . . . 162
6.24 Total pressure contours for water flow at P set =0.6 MPa. . . . . . . . . . . 163
6.25 Influence of cavitation on flowforce versus valve opening . . . . . . . . . 164
6.26 Density contours for air flow at P set =0.6 MPa. Solid white thick line
indicated sonic position. . . . . . . . . . . . . . . . . . . . . . . . . . . 165
6.27 Experimental-CFD flowforce for air and water at P set =0.3 MPa. . . . . . 166
6.28 Influence of adjustment ring location on disk flowforce. . . . . . . . . . . 167
6.29 Pressure drop coefficient ζ at full lift and L=5.5 mm versus volumetric
quality for P set =0.15 MPaG. . . . . . . . . . . . . . . . . . . . . . . . . 169
6.30 Verification of independence of pressure drop coefficient on set pressure. . 169
6.31 Dimensionless pressure drop ΦL function of volumetric quality of air at
full lift for P set =0.15 and 0.3 MPaG. Comparison with Chisholm (1971)
correlation. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 170
6.32 3D plot of β-L-ζ of transparent SRV at P set =0.15 MPaG. . . . . . . . . . 170
6.33 Critical mass flux measured and calculated with ω-method for non-flashing
conditions in transparent SRV. . . . . . . . . . . . . . . . . . . . . . . . 171
xi
xii List of figures
6.34 Measured and calculated mass flux with ω-method for non-flashing con-
ditions under various set pressures in metallic valve. . . . . . . . . . . . . 172
6.35 Two-phase discharge coefficient versus volumetric quality of air for trans-
parent and industrial valve. Comparison with Lenzing et al. (1998) formula.175
6.36 Two-phase discharge coefficient versus valve lift in single and two-phase
flow. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 176
6.37 Experimental and predicted from ω-method, API520 and HNE-DS dis-
charged mass flux at full lift for P set =0.15 MPa in transparent valve. . . . 176
6.38 Experimental and predicted from ω-method, API520 and HNE-DS dis-
charged mass flux at full lift for P set =0.3 MPa in transparent valve. . . . . 177
6.39 Calculated versus measured mass flux for different set pressures in indus-
trial valve. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 178
6.40 Relative discrepancy between predicted and measured pressure drop co-
efficient in SRV. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 179
B.1 Detailed optical probe test section. . . . . . . . . . . . . . . . . . . . . . 204

B.2 Transparent SRV test section AGATHE II. . . . . . . . . . . . . . . . . 205
B.3 LUCY III test section drawing. . . . . . . . . . . . . . . . . . . . . . . 206
Force sensor mounted in safety valve WEIR of type 1 1/2 G 3 . . . .
00 00
B.4 . . 207
C.1 Three Mile Island nuclear power plant. . . . . . . . . . . . . . . . . . . 210

C.2 Accident mainly caused by missing safety valves (Mexico city 1984). . . 211
D.1 Diagrams 1-4 of Annex II in PED . . . . . . . . . . . . . . . . . . . . . 214

D.2 Diagrams 5-8 of Annex II in PED . . . . . . . . . . . . . . . . . . . . . 215
D.3 Diagram 9 of Annex II in PED: Piping referred to in Article 3, Section
1.3(b), second indent. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 216
F.1 Vertical and horizontal evaporation-condensation in a tube (from Collier

and Thome (1994)). . . . . . . . . . . . . . . . . . . . . . . . . . . . . 220
G.1 Flow patterns in vertical upward flow in a pipe. Adapted from Weisman
(1983). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 222
G.2 Vertical upward flow map of Hewitt and Roberts (1969). . . . . . . . . . 224
G.3 Flow regimes identified in vertical downward cocurrent flow taken from
Kourakos et al. (2007). . . . . . . . . . . . . . . . . . . . . . . . . . . . 225
G.4 Vertical downward flow map of Golan and Stenning (1969). . . . . . . . 226
G.5 Flow map in horizontal and slightly inclined pipe proposed by Taitel and
Dukler (1977) with the characteristic numbers K, F and T as a function of
the Lockhart and Martinelli (1949) parameter. . . . . . . . . . . . . . . . 227
H.1 Moody (1944) diagram. . . . . . . . . . . . . . . . . . . . . . . . . . . 231
I.1 Two-phase multiplier Φ2L0 for different mass qualities and pressures given
by Martinelli and Nelson (1948). . . . . . . . . . . . . . . . . . . . . . 234
I.2 Void fraction versus mass quality for different pressures by Martinelli and
Nelson (1948). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 235
xii
List of figures xiii
I.3 Acceleration multiplier r2 (left), gravitational multiplier r4 (right up) and

frictional multiplier r3 (right down) function of the operating pressure for
different mass qualities given by Thom (1964). . . . . . . . . . . . . . . 237
I.4 Baroczy (1966) two-phase flow multiplier. . . . . . . . . . . . . . . . . 238
I.5 Baroczy (1966) curves for mass flux correction factor. . . . . . . . . . . 239
I.6 Comparison of different two-phase flow models (Todreas and Kazami
(1989)). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 240
J.1 Different types of singularities and the proposed pressure drop coefficients
by (Idel’Cik (1986)). . . . . . . . . . . . . . . . . . . . . . . . . . . . . 243
J.2 Demonstration of the pressure changes in incompressible fluid at abrupt
contraction and expansion geometries (taken from Todreas and Kazami
(1989)). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 245
J.3 Simplified schematic of pressure drop in LUCY II. . . . . . . . . . . . . 245
xiii
List of Tables
2.1 Classification of relief devices Tyco (2008). . . . . . . . . . . . . . . . . 37

2.2 Units for two-phase SRV methodology. . . . . . . . . . . . . . . . . . . 46
2.3 Summarizing matrix of two-phase flow methodology in SRV. . . . . . . 54
2.4 Discharge coefficient Kd proposed by Boccardi et al. (2005). . . . . . . . 56
3.1 Presentation of the four main experimental facilities. . . . . . . . . . . . 65

3.2 Summarizing matrix of experimental tests. . . . . . . . . . . . . . . . . 66
3.3 Test matrix for pressure, optical probe measurements and flow visualiza-
tion. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70
3.4 Summarizing matrix of different test cases studied. . . . . . . . . . . . . 74
3.5 Upstream conditions for pressure measurements. . . . . . . . . . . . . . 74
3.6 Upstream conditions for optical probe measurements and flow visualization. 74
3.7 Position of different pressure taps along the radius of the valve disk. . . . 77
3.8 Test matrix for SRV study. . . . . . . . . . . . . . . . . . . . . . . . . . 83
3.9 Optical probe and flow visualization tests in AGATHE II setup. . . . . . . 85
4.1 Repeatability test for optical probe measurement (taken from Fernandes
et al. (2010)). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 120
4.2 Proposed correlation for different expansion singularities-regression pa-
rameters and average deviation from experimental measurements. . . . . 122
5.1 Proposed correlation for different contraction singularities-regression pa-

rameters and average deviation from experimental measurements. . . . . 138
6.1 Flow conditions for SRV tests. . . . . . . . . . . . . . . . . . . . . . . . 161
E.1 Relief Valve Orifice Sizes . . . . . . . . . . . . . . . . . . . . . . . . . . 217
I.1 Values of coefficient K and C0 for drift flux model proposed by several
authors. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 236
I.2 Comparison of two-phase pressure drop correlations for steam-water mix-
tures by Idsinga et al. (1977). . . . . . . . . . . . . . . . . . . . . . . . . 240
I.3 Comparison of flow regime dependent models by Beattie (1973) (taken
frrom Delhaye et al. (1980)). . . . . . . . . . . . . . . . . . . . . . . . . 241
J.1 Pressure loss coefficient range for different singularities (taken from To-
dreas and Kazami (1989)) . . . . . . . . . . . . . . . . . . . . . . . . . . 244
xvi List of tables
J.2 Pressure changes in abrupt contraction and expansion geometries in in-

compressible and compressible fluid (taken from Todreas and Kazami
(1989)). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 246
K.1 Results in sudden expansion σ=0.43. . . . . . . . . . . . . . . . . . . . 249

K.2 Results in sudden expansion σ=0.65. . . . . . . . . . . . . . . . . . . . 250
K.3 Results in smooth expansion σ=0.43, α=5 ◦ . . . . . . . . . . . . . . . . 251
K.6 Deviation experimental-literature correlation for expansion. . . . . . . . 254
K.7 Results in sudden contraction of σ=2.34. . . . . . . . . . . . . . . . . . 257
K.8 Results in smooth contraction of σ=2.34, α=15 ◦ . . . . . . . . . . . . . . 258
K.9 Results in smooth contraction of σ=1.56, α=9 ◦ . . . . . . . . . . . . . . 259
K.10 Deviation experimental-literature correlation for contraction. . . . . . . . 260
K.11 Results in transparent SRV at P set =0.15 MPa. . . . . . . . . . . . . . . . 261
xvi
List of Symbols
Q Volumetric flow rate [l/s] . . . . . . . . . . . . . . . . . . . . . . 15

A Area [m2 ] . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
M Mass flow rate [kg/s] . . . . . . . . . . . . . . . . . . . . . . . . 15
X Phase density function [-] . . . . . . . . . . . . . . . . . . . . . . 16
t Time [s] . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
αlocal Local void fraction [-] . . . . . . . . . . . . . . . . . . . . . . . . 16
α Void fraction [-] . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
β Volumetric quality of air [-] . . . . . . . . . . . . . . . . . . . . . 16
G Mass flux [kgm−2 s−1 ] . . . . . . . . . . . . . . . . . . . . . . . . 17
x Mass quality of air [-] . . . . . . . . . . . . . . . . . . . . . . . . 17
J Superficial velocity [ms−1 ] . . . . . . . . . . . . . . . . . . . . . 17
UL Average liquid velocity [ms−1 ] . . . . . . . . . . . . . . . . . . . 17
UG Average gas velocity [ms−1 ] . . . . . . . . . . . . . . . . . . . . 17
α slip Void fraction with slip velocity [-] . . . . . . . . . . . . . . . . . 17
U∗ Two-phase average velocity of the mixture [ms−1 ] . . . . . . . . . 18
ρ∗ Homogeneous flow density [kgm−3 ] . . . . . . . . . . . . . . . . 18
ρmix Separated flow mixture density [kgm−3 ] . . . . . . . . . . . . . . 18
σ Surface tension [kgs−2 ] . . . . . . . . . . . . . . . . . . . . . . . 21
µ Dynamic viscosity [kg (ms)−1 ] . . . . . . . . . . . . . . . . . . . 21
P Pressure [Pa] . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
Φ Two-phase flow multiplier [-] . . . . . . . . . . . . . . . . . . . . 25
k Pressure loss coefficient [-] . . . . . . . . . . . . . . . . . . . . . 27
G Mass flux [kgm−2 σ−1 ] . . . . . . . . . . . . . . . . . . . . . . . . 28
vL Liquid specific volume [m3 kg−1 ] . . . . . . . . . . . . . . . . . . 28
vG Gas specific volume [m3 kg−1 ] . . . . . . . . . . . . . . . . . . . 28
CC Contraction coefficient [-] . . . . . . . . . . . . . . . . . . . . . . 30
τ Wall shear stress [Pa] . . . . . . . . . . . . . . . . . . . . . . . . 30
B Parameter of Chisholm (1971) [-] . . . . . . . . . . . . . . . . . . 33
P set Set pressure [MPa] . . . . . . . . . . . . . . . . . . . . . . . . . 36
Pc Closing pressure [MPa] . . . . . . . . . . . . . . . . . . . . . . . 36
Pb Built-up back pressure [MPa] . . . . . . . . . . . . . . . . . . . . 36
Kd Discharge coefficient [-] . . . . . . . . . . . . . . . . . . . . . . . 36
0
Qm Experimental discharged flow rate [kg/h] . . . . . . . . . . . . . . 39
Qm Theoretical discharged flow rate [kg/h] . . . . . . . . . . . . . . . 39
Kdr Discharge coefficient after exhaustion [-] . . . . . . . . . . . . . . 40
k Ratio of specific heats [-] . . . . . . . . . . . . . . . . . . . . . . 40
xviii List of symbols
CP Specific heat capacity of a gas for constant pressure [-] . . . . . . 40

CV Specific heat capacity of a gas for constant volume [-] . . . . . . . 40
M Molar mass [kgkmol−1 ] . . . . . . . . . . . . . . . . . . . . . . . 40
Z Compressibility factor [-] . . . . . . . . . . . . . . . . . . . . . . 40
σ0 Cavitation number [-] . . . . . . . . . . . . . . . . . . . . . . . . 41
c Speed of sound [m/s] . . . . . . . . . . . . . . . . . . . . . . . . 42
K Bulk modulus of elasticity of the fluid [Pa] . . . . . . . . . . . . . 42
cmix Speed of sound in mixture medium [ms−1 ] . . . . . . . . . . . . . 43
q Heat absorbed by surrounding in thermodynamic system [J] . . . 47
H Enthalpy of vaporization [J] . . . . . . . . . . . . . . . . . . . . 47
WS Work done on the surrounding of thermodynamic system [J] . . . 47
S Entropy [JK−1 ] . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
P0 Stagnation pressure [Pa] . . . . . . . . . . . . . . . . . . . . . . 47
P1 Throating pressure [Pa] . . . . . . . . . . . . . . . . . . . . . . . 47
L Latent heat of vaporization [Jmol−1 ] . . . . . . . . . . . . . . . . 48
C PL Liquid specific heat at constant pressure [JKg−1 K−1 ] . . . . . . . . 48
v Specific volume [m3 kg−1 ] . . . . . . . . . . . . . . . . . . . . . . 50
ncrit Critical pressure ratio Pcrit /P1 [-] . . . . . . . . . . . . . . . . . . 50
n1 Back pressure ratio P1 /P0 [-] . . . . . . . . . . . . . . . . . . . . 50
N Boiling delay factor Diener and Schmidt (2004) [-] . . . . . . . . 50
ωDS ω parameter of Diener and Schmidt (2004) [-] . . . . . . . . . . . 52
α Diener and Schmidt (2004) parameter [-] . . . . . . . . . . . . . . 52
Kb Correction factor due to back pressure for gas [-] . . . . . . . . . 53
R Gas constant [Pa·m3 K−1 kg−1 mol−1 ] . . . . . . . . . . . . . . . . . 53
Z Compressibility factor (Z=1 for ideal gas) [-] . . . . . . . . . . . 53
KW Correction factor due to back pressure for liquid [-] . . . . . . . . 53
KV Correction factor due to viscosity [-] . . . . . . . . . . . . . . . . 53
m, q, r2 Regression parameters of correlation by Boccardi et al. (2005) [-] . 56
h Valve lift [mm] . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
Lnozzle Length of inlet nozzle of PRV [m] . . . . . . . . . . . . . . . . . 57
k Turbulent kinetic energy [kgm−1 s−3 ] . . . . . . . . . . . . . . . . 58
ω Specific turbulence dissipation rate [kgm−3 s−2 ] . . . . . . . . . . . 58
CV Flow coefficient [-] . . . . . . . . . . . . . . . . . . . . . . . . . 58
d1 Upstream diameter [mm] . . . . . . . . . . . . . . . . . . . . . . 69
D2 Downstream diameter [mm] . . . . . . . . . . . . . . . . . . . . 69
fsamp Sampling frequency [Hz] . . . . . . . . . . . . . . . . . . . . . . 73
tacq Acquisition time [s] . . . . . . . . . . . . . . . . . . . . . . . . . 73
Ls Length of singularity [mm] . . . . . . . . . . . . . . . . . . . . . 74
F Force on valve disk [N] . . . . . . . . . . . . . . . . . . . . . . . 76
R Radius of the valve disk [mm] . . . . . . . . . . . . . . . . . . . 76
P (r) Pressure along valve disk radius [MPa] . . . . . . . . . . . . . . . 77
B Magnetic flux density [T] . . . . . . . . . . . . . . . . . . . . . . 85
E Signal voltage [V] . . . . . . . . . . . . . . . . . . . . . . . . . . 86
n Refractive index [-] . . . . . . . . . . . . . . . . . . . . . . . . . 87
T gas Time of gas meeting the probe [s] . . . . . . . . . . . . . . . . . 88
T acq Time of probe acquisition [s] . . . . . . . . . . . . . . . . . . . . 88
xviii
List of symbols xix
T f light Transient time between primary and secondary sensor [ms] . . . . 89

Vb Bubble velocity [ms−1 ] . . . . . . . . . . . . . . . . . . . . . . . 89
h(l) Chord length distribution [-] . . . . . . . . . . . . . . . . . . . . 90
g(l) Bubble diameter distribution [-] . . . . . . . . . . . . . . . . . . 90
fint Bubble interference frequency [s−1 ] . . . . . . . . . . . . . . . . 90
Nb Total number of bubbles passing from the probe [-] . . . . . . . . 90
Ai Interfacial area concentration [m−1 ] . . . . . . . . . . . . . . . . . 90
D sm Sauter mean diameter [mm] . . . . . . . . . . . . . . . . . . . . . 90
∆P sing Singular pressure change [kPa] . . . . . . . . . . . . . . . . . . . 95
∆Preg−meas Regular measured pressure change [kPa] . . . . . . . . . . . . . . 95
∆P sing−meas Singular measured pressure change [kPa] . . . . . . . . . . . . . 95
∆Preg−calc Regular calculated pressure change [kPa] . . . . . . . . . . . . . 95
L/d Reattachment length [-] . . . . . . . . . . . . . . . . . . . . . . . 95
ΦLst Dimensionless static singular pressure change [-] . . . . . . . . . 97
α0 Expansion opening angle corresponding to the strongest pressure
recovery for a diffuser (Comolet (1963)) . . . . . . . . . . . . . . 100
ΦL Two-phase multiplier [-] . . . . . . . . . . . . . . . . . . . . . . 103
ζL Liquid pressure drop coefficient [-] . . . . . . . . . . . . . . . . . 103
ζT P Two-phase mixture pressure drop coefficient [-] . . . . . . . . . . 103
D32 Sauter mean diameter [mm] . . . . . . . . . . . . . . . . . . . . . 114
D peak Peak diameter in the bubble size distribution diagram [mm] . . . . 115
mb Mass of bubble [kg] . . . . . . . . . . . . . . . . . . . . . . . . . 117
Vb Bubble volume [m3 ] . . . . . . . . . . . . . . . . . . . . . . . . . 117
CD Drag coefficient [-] . . . . . . . . . . . . . . . . . . . . . . . . . 117
s Dimensionless slip velocity [-] . . . . . . . . . . . . . . . . . . . 150
m Mass of plate [kg] . . . . . . . . . . . . . . . . . . . . . . . . . . 152
∆F Discrepancy between flowforce in two-phase flow and water for
SRV [%] . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 158
FT P Flowforce in two-phase flow for SRV [N] . . . . . . . . . . . . . 158
FS P,water Flowforce in single-phase water flow for SRV [N] . . . . . . . . . 158
Ma Mach number [-] . . . . . . . . . . . . . . . . . . . . . . . . . . 163
ṁexp Experimental mass flow rate through SRV [kg/s] . . . . . . . . . . 172
ṁnozzle Mass flow rate through isentropic ideal nozzle [kg/s] . . . . . . . 172
α Boiling delay exponent [-] . . . . . . . . . . . . . . . . . . . . . 174
ρ Density [kgm−3 ] . . . . . . . . . . . . . . . . . . . . . . . . . . . 223
F Taitel and Dukler (1977) non-dimensional parameter (wavy annu-
lar & wavy-intermittent transition) [-] . . . . . . . . . . . . . . . 226
K Taitel and Dukler (1977) non-dimensional parameter (stratified to
wavy flow transition) [-] . . . . . . . . . . . . . . . . . . . . . . 226
Θ Inclination angle of the pipe in respect to the horizontal plane and
with a positive orientation the vertical descending flow (Taitel and
Dukler (1977))[ ◦ ] . . . . . . . . . . . . . . . . . . . . . . . . . . 226
T Taitel and Dukler (1977) non-dimensional parameter (bubbly to
intermittent/plug transition) [-] . . . . . . . . . . . . . . . . . . . 226
λ Darcy friction factor [-] . . . . . . . . . . . . . . . . . . . . . . . 229
L Pipe length [m] . . . . . . . . . . . . . . . . . . . . . . . . . . . 229
xix
xx List of symbols
D Large pipe diameter [m] . . . . . . . . . . . . . . . . . . . . . . 229

U Velocity [ms−1 ] . . . . . . . . . . . . . . . . . . . . . . . . . . . 229
Re Reynolds number [-] . . . . . . . . . . . . . . . . . . . . . . . . 230
ν Kinematic viscosity [m2 s−1 ] . . . . . . . . . . . . . . . . . . . . 230
k Roughness of the pipe [m] . . . . . . . . . . . . . . . . . . . . . 230
r4 Length-averaged two-phase gravitational pressure drop multiplier
[-] . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 234
r2 Length-averaged two-phase pressure drop multiplier due to accel-
eration [-] . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 234
r3 Length-averaged two-phase frictional pressure drop multiplier [-] . 234
C0 Concentration parameter [-] . . . . . . . . . . . . . . . . . . . . . 235
Roughness of the pipe [mm] . . . . . . . . . . . . . . . . . . . . 241
We Weber number [-] . . . . . . . . . . . . . . . . . . . . . . . . . . 241
g Gravitational acceleration [ms−2 ] . . . . . . . . . . . . . . . . . . 245
d Small pipe diameter [m] . . . . . . . . . . . . . . . . . . . . . . 246
Jt1 Total upstream velocity [ms−1 ] . . . . . . . . . . . . . . . . . . . 249
xx
Abbreviations
LM Lockhart and Martinelli (1949) parameter

SP Single-Phase
TP Two-Phase
BWR Boiling-Water Reactor
PWR Pressurized-Water Reactor
LWR Light-Water Reactor (PWR or BWR)
LOCA Loss of Coolant Accident
RD Relief Device
RV Relief Valve
SV Safety Valve
PRV Pressure Relief Valve
SRV Safety Relief Valve
PORV Pilot Operated Relief Valve
LPG Liquefied Petroleum Gas
LNG Liquefied Natural Gas
HP High Pressure
HEM Homogeneous Equilibrium Model
HNE Homogeneous Non-Equilibrium Model
HNE-DS HNE Diener and Schmidt (2004) model
VKI von Karman Institute for Fluid Dynamics
CETIM Centre Technique des Industries Mécaniques
ULB Université Libre de Bruxelles
KIT Karlsruhe Institute of Technology
API American Petroleum Institute
RP Recommended Practice
ISO International Organization for Standardization
NF Norme Française
DIN Deutsches Institut für Normung
AIChE American Institute of Chemical Engineers
DIERS Design Institute for Emergency Relief Systems
ASME American Society of Mechanical Engineers
NRC Nuclear Regulatory Commission
BLEVE Boiling Liquid Expanding Vapor Explosion
ID Inner Diameter
DN Nominal Diameter
PN Nominal Pressure rating
xxii Abbreviations
ECT Electrical Capacitance Tomography

WMS Wire Mesh Sensor Tomography
FEM Finite Element Method
CFD Computational Fluid Dynamics
RANS Reynolds-Averaged Navier-Stokes
SST Shear Stress Transport
MSWP Maximum Safe Working Pressure
FS Full Scale
PMMA Polymethyl Methacrylate
BEP Best Efficiency Point
PDF Probability Density Function
PED Pressure Equipment Directive
EU European Union
EC European Commission
PS Maximum Working Pressure
BP British Petroleum
CF Cavitating Flow
xxii
Part I
Preface
1 Fundamentals of the study
1.1 Introduction
1.1.1 Basics of two-phase flows
Two-phase flow is the area of fluid mechanics that describes the flow of mixtures consist-
ing of two or more immiscible phases. This type of flow is the simplest case of multiphase
flow. The different phases of multiphase flow are liquid, gas and solid. Two-phase flow
is constantly met in our daily practice. Sandstorm, fog, snow and rain are some natural
examples. In industrial processes, several examples of such flow can also be found. Over
half of all chemical engineering is concerned with multiphase flow. Air and water two-
phase flow is used in water treatment processes. Well known examples are the water-air
backwashing of rapid sand filters and the air-water scouring of pipelines in the distribution
network.
Gas-liquid flows have been studied extensively for many years, largely because of
their immense frequency of occurrence throughout industry. Two-phase flow arises when
the averaged motion of one material (or phase, field, fluid or component) is distinctly
different from that of another. The goal is to predict the averaged behavior of a two-
phase flow field with a theory or model that is rather general and useful. This type of
flow involves the combined flow of a liquid and a gas or vapor phase. It is a difficult
subject to investigate principally because of the complexity of the form in which the two
fluids exist inside the pipe different components, known as the flow regime. Building
a model from first principles in all but the most elementary situations is a complicated
issue. Dimensional analysis is used to establish the relevant groups to aid in designing the
suitable experiments. Most available empirical results are applicable only to gas-liquid
two-phase flow.
For several industrial purposes, it is important to predict the pressure loss occurring
in two-phase flow, as well as the void fraction, which controls the effective density of
the mixture of gas and liquid. Since the gas and liquid may move with different average
velocities in a channel, the void fraction (or hold-up) depends on the flow rates of each
phase as well as on the relative velocity between the phases. The liquid hold-up for
instance is not necessarily equal to the relative fraction of that phase in the entering fluid
mixture. As the average velocity of the gas increases compared to the liquid, the void
fraction decreases due to the shorter average residence time of the gas phase. The velocity
difference between phases depends on how well the phases are coupled, which depends in
turn on the amount of surface area available for the transfer of momentum between them.
The difference in the velocity of the two phases is called slip ratio, or simpler just slip and
4 1 Fundamentals of the study
in the case of their homogeneous mixing this ratio is equal to unity.

Analysis of two-phase flows involves prediction of the flow pattern, the liquid-vapor
density, pressure drop across a given channel length, flow stability, maximum flow rates,
and heat transfer rates. The behavior of two-phase flows can be quite complicated and
can strongly depend on the relative flow rates of the two-phases, channel orientation,
singularities, fluid properties and inlet conditions. The orientation of the pipe makes
a difference in the flow regime because of the role played by gravity and the density
difference between the two fluids. The usual question for the engineer is that of calculating
the pressure drop required to achieve specified flow rates of the gas and the liquid through
a pipe of a given diameter.
In many applications, self-vaporization of liquid can occur due to pressure or temper-
ature change (such as for pressurized water). In this occurrence, vapor-liquid mixture is
created and it is very important to recognize and be able to calculate when these phases
are in equilibrium with each other and the fraction of the circuit occupied by each phase.
A typical example of vapor-liquid incidence is in core of nuclear reactors where water un-
der high pressure is used as coolant to increase the boiling point. Hence, often two-phase
mixtures are accidentally or deliberately produced. Thus, controlling the flow in this cir-
cumstances is of vital importance for nuclear safety. Two-phase flow can also take place
when sudden depressurization of liquid happens and vapor bubbles are generated; when
this phenomenon happens partially is called cavitation while when there is a complete
transformation from liquid to gas is called flashing flow.
This research project is devoted to air-water flow investigation simulating the behavior
of liquid-vapor although such mixtures are not considered. Geometrical singularities are
tested in the horizontal plane in single and two-phase flow. During the study of a safety
relief valve, in addition to single and two-phase flow, cavitation phenomena are observed.
Combined vertical-horizontal flow (change of flow direction) takes place in safety valve
configuration.
1.1.2 Two-phase flow occurring in industrial devices and safety valves

The use of singular geometries is found in pipings and fittings with large and restricted
flow area and their combination. As for example the case at the inlet and exit of the fuel
assemblies. Additionally, most boiling reactors include regions where sudden expansions
occur. Moreover, the design of many heat exchangers using horizontal tubular elements
necessitates that these elements be interconnected using multiple return bends design to
form a serpentine arrangement. Frequently, two-phase flow can occur either accidentally
or on purpose and the geometrical discontinuity enhances the complexity of the phenom-
ena. Examples of pipe fittings are illustrated in Fig. 1.1.
Pressure safety valves are used in the oil and gas production,petrochemical and fine
chemical industries as well as in nuclear reactor protection. Additionally, use of such
devices is mandatory in water supply networks. Applications involving two-phase flows
range from straightforward transfer systems, such as pumping, to those involving heat
and/or mass transfer, such as heat exchangers, boilers, cooling towers, bubble columns,
fluidized beds and nuclear reactors. The combined flow of a liquid and a gas or va-
por phase occurs in the normal operation of Boiling Water Reactors (BWRs), as well as
on the secondary loop of the steam generators of Pressurized Water Reactors (PWRs).
4
1.1 Introduction 5
Figure 1.1: Different types of pipe fittings-geometrical singularities (SS Engineers & Con-
sultants (2010)).
Two-phase flow also plays a major role in many light-water reactor (LWR) transients and
accidents as for instance the loss of coolant accident (LOCA). An example of a nuclear
reactor core is shown in Fig. 1.2.
Hereunder various applications of safety valves are summarized:
Nuclear applications:
Figure 1.3 shows a simplified schematic of the steam generator and components typi-
cal of a large turbine unit. The position at which the pressure relief, control and emergency
stop valves are placed is indicated.
Safety valves installed on top of the boiler protect the steam system components from
over pressure. The pressure from the boilers drives the steam to the high pressure (HP)
turbine. On route to the turbine the steam travels through several valves. Two of interest
are the emergency stop valves and the governor valves. The governor valve controls the
quantity of steam flowing to the turbine, and therefore controls the speed of the turbine
when not connected to the grid, and when the generator is synchronized to the grid, it
determines the electrical output of the unit. Before reaching the governor valve the steam
passes through the emergency stop valve. The emergency stop valve quickly stops the
steam flow to the turbine in the event of an emergency that could damage the turbine
(Bereznai (2005)).
Chemical applications:
Two-phase flow can be met in refrigerant systems, heat pumps and polymerization
reactors. Pressure relief valves are used in petroleum refining, petrochemical and chem-
ical manufacturing, natural gas processing and power generation industries. This device
is also used in the Bayer process; the principal industrial means of refining bauxite to
5
Figure 1.2: Detailed sketch of nuclear reactor core.
Figure 1.3: Schematic of the steam system and components typical of a large turbine unit
(Taken by Bereznai (2005)).
produce alumina.
Cryogenic applications:
Within a cryogenic system, adequate relief valves must be installed for all vacuum
and cryogenic vessels, and also for any cryogenic lines that have the potential to trap
cryogenic fluids. Relief valves must be sized so that under worst-case failure conditions,
the maximum pressure reached in any vessel is below the maximum safe working pressure
(MSWP) for the vessel. No fixed prescription can be given to determine valve sizing for
6
1.2 Objectives-Achievements / Motivation 7
all, or even most cases. Each system must be analyzed in detail to properly determine
worst-case failure modes and the required relief valve sizing.
Chemical reactors containing compressed liquefied gases such as Chlorine, Ammonia,
Propane, Liquefied Natural Gas (LNG) and Propane in large pressure vessels are protected
from overpressure with Pressure Relief Valves. Processing, storage and transportation of
theses reservoirs should be handled with care to avoid accidents.
Examples of safety valves manufactured by FLOW SAFE, Inc. in cryogenic instal-
lations are shown in Fig 1.4. These valves were tested at the Cryogenics Testbed at the
Kennedy Space Center, operated by the Dynacs Engineering Co.
(a) Pilot Operated Safety Relief Valve during (b) Spring Operated Safety Relief Valve during
cryogenic flow testing at Kennedy Space Center cryogenic flow testing at Kennedy Space Center
Figure 1.4: Safety Relief Valves tested under cryogenic conditions in Kennedy Space
Center.
1.2 Objectives-Achievements / Motivation

In many chemical and nuclear processes pressure relief valves (PRV) are used. A PRV is
a device used to prevent the excessive rising of the pressure in the pressurized reservoir,
process or system in general. This is accomplished by discharging a certain amount of
fluid. The use of these devices is very crucial for the industry. Indeed, these simple
and robust in their design valves are the ultimate protection when all other systems are
insufficient. The poor design of these devices can be disastrous for the environment or
even, in the worst case, it can cause injuries or deaths (Chabane et al. (2009)).
A cutaway drawing of a PRV commonly met in the industry is presented in Fig. 1.5.
The nozzle, valve body, spring, valve disk and downstream section can be distinguished.
Further analysis and more details on the role of each component of the valve is given in
the next chapters. For illustration purposes, a schematic of the problem to be investigated
is depicted in Fig. 1.6; a pressurized tank, fed from a primary circuit through a control
valve, is protected with a safety relief valve. When a predetermined pressure is reached,
fluid is discharged through the valve with the aim of dropping the pressure to acceptable
levels. The purpose for engineers is to size correctly the valve. The question that arises is
whether the design of the valve has to be modified for the use of the same valve when the
latter functions in two-phase flow regime.
7
Therefore, the main objective of this study is to determine whether a two-

phase mixture passing through a PRV approaches more to compressible or
incompressible fluid behavior and for which range of conditions this happens.
The latter information could be useful for PRV manufacturers to improve
valve modeling in two-phase flow operation.
After a thorough literature survey, a relatively pronounced research lack in this par-
ticular field can be stated. In the present study, injected air-water mixture with cavitating
flow through SRV is investigated. Most references are focused in water-vapor mixtures
with flashing or non-flashing flow, the latter proving the originality of this thesis. A ho-
mogeneous mixture approach with extension of Omega-methodology is applied by mod-
ification of ω parameter for the case of air-water flow with cavitation.
Additionally, in the frame of this project an innovative transparent SRV facility has
been built allowing for flow visualization in single-phase flow (cavitation bubbles) and
injected air-water flow (for very low air fractions). Moreover, one of the most important
contributions of this project was the study of the valve opening characteristics by mea-
suring the flowforce (especially at the lowest disk lifts) in SRV for incompressible and
compressible flow. The former have been compared with two-phase flow revealing the
compressible behavior of these mixtures above certain flow conditions and geometrical
stipulation. Furthermore, a relatively simple 2D CFD model that could predict with good
accuracy the flowforce applied in SRV for compressible or incompressible flow has been
developed. For this model, locally (close to the valve disk) axisymmetric flow is assumed.
With the use of these results, the role of adjustment ring on flowforce has been quantified.
Finally, several geometrical singularities have been studied in order to obtain infor-
mation about possible simplified modeling of SRV with as a series of expansion and/or
contraction. Pressure evolution along each singularity permitted establishing the pressure
drop. Flow visualization in selected divergence section was performed and a flow pattern
map for this specific geometry is specified. The uniqueness of this map is found in its form
since the input is the flow condition before divergence while output is the flow regime af-
ter the expansion. Pressure drop correlations are deduced for the case of dispersed bubbly
flow and no flow regime transition in each geometry. Thus, due to unknown flow pattern
transition in the valve, these results could not be exploited for SRV modeling. Likewise,
an exclusivity in this study is the discussion of the influence of opening angle on two-
phase pressure drop in divergence sections.
Summarizing the main goals and original achievements of the thesis:
I Flow regime dependent ∆P correlations for each singularity studied are pro-
posed.
I Flow pattern map with inlet upstream conditions and outlet downstream flow
regime are established; flow structure is also identified with optical probe
measurements.
I Transparent SRV facility on selected SRV is built allowing flow visualization
of cavitation and/or injected air-water flow.
I Study of air-water flow with cavitation in SRV and modification of HNE-DS
applied to cavitating flows (CF).
8
1.2 Objectives-Achievements / Motivation 9
I Measurements of flowforce in SP and TP flow revealing the compressible

fluid character of TP mixtures above certain limits.
I Simple 2D axisymmetric CFD study of a SRV for prediction of flowforce in
compressible and incompressible flow.
Soupape de decharge
Introduction
Safety valve Qm,out
Control valve
Pressurized tank
Qm,in
Figure 1.5: Cutaway of PRV. Figure 1.6: Schematic of the problem.
An example of a recent catastrophe that took place in the Gulf of Mexico and for
which one possible cause is linked to a wrongly plumbed PRV is shown in Fig. 1.7. More
details about the aforementioned and other accidents that have been reported and were
related to PRVs are given in Appendix C. Hence, we can conclude that the use of PRV
in industrial safety is of vital importance since their absence, wrong sizing or incorrect
installation can cause hazardous situations.
Figure 1.7: Deepwater Horizon oil spill accident (2010).
The Pressure Equipment Directive (PED)

The Pressure Equipment Directive 97/23/EC PED (1997) (called PED) was formulated
under the “New Approach to Technical Harmonization and Standards” and aims at har-
monizing the national provisions, specific to each member state, in order to enable free
Deepwater Horizon oil spill (2010)
9
flow of equipment under pressure within the European Union. It sets out the standards for
the design and fabrication of pressure equipment (meaning vessels, piping, safety valves
and other components and assemblies subject to pressure loading) generally over 1 liter
in volume and having a maximum pressure (PS) equal or higher than 0.5 barg relative
pressure. It also sets the administrative procedures requirements for the “conformity as-
sessment” of pressure equipment, for the free placing on the European market without
local legislative barriers.
This Directive was adopted by the European Parliament and the European Council in
May 1997. It has initially come into force on 29 November 1999. From that date until 28
May 2002 manufacturers had a choice between applying the pressure equipment directive
or continuing with the application of the existing national legislation. Since 29 May 2002
the pressure equipment directive is mandatory throughout the EU.
Typical equipment that is exposed to PED consists of vessels, pressurized storage
containers, heat exchangers, steam generators, boilers, industrial piping, safety devices
and pressure accessories. These applications are widely used in the process industries (oil
& gas, chemical, pharmaceutical, plastics and rubber and the food and beverage industry),
high temperature process industry (glass, paper and board), energy production and in the
supply of utilities, heating, air conditioning, gas storage and transportation.
The directive provides, together with the directives related to simple pressure ves-
sels (2009/105/EC), transportable pressure equipment (99/36/EC) and Aerosol Dispensers
(75/324/EEC), an adequate legislative framework on European level for equipment sub-
ject to a pressure hazard.
Classification of equipment under pressure (Endress+Hauser (2010)):

The equipment under pressure is classified according to the potential risk in categories
from I to IV. The potential danger depends from one part on the product of the maximum
allowance pressure from either the pressurized volume, either from the nominal diameter
of piping and from other part of the type of the fluid being under pressure and from their
dangerousness.
Two kinds of fluids should be considered:
Gases
Gas, liquefied gas, dissolved gas under pressure, vapor as well as liquids with vapor pres-
sure higher than 0.05 MPa relative to the maximum admissible pressure.
Liquids
Liquids with vapor pressure lower or equal than 0.05 MPa relative to the maximum ad-
missible pressure.
Every type of fluid can be divided in two groups:

Group 1: Dangerous fluids: explosives, oxidants, toxic or inflammables.
Group 2: Harmless fluids: nitrogen, air, water vapor.
The distinction of equipment under pressure in categories is made in accordance with
diagrams 1-9 of Annex II of PED (see Appendix D). Safety valves belong to category IV
of safety accessories.
The requirement of protecting equipment under pressure has placed safety valves on
the front line for protection of such components.
10
1.3 Methodology 11
1.3 Methodology
The methodology followed during this project is split in two phases. At first, several
geometrical singularities are studied in terms of pressure evolution characterization and
flow rate variation. A uniform two-phase flow mixture is created upstream the singularity
by injecting air through small orifices in pure water flow. The flow structure is identified
by flow visualization performed with a high-speed camera. The void fraction, bubble
size and velocity are determined with a dual optical probe. Measurements are acquired
upstream and downstream at different positions. In Fig. 1.8 a schematic of the procedure
followed during singularities study is illustrated.
Geometrical singularity
ζ Æ ∆P correlations
∆Psingular
P1
Flow rate P0
measurement
Gas injector
Void fraction, bubble size

and velocity
Flow pattern change?
(Flow map)
Figure 1.8: Methodology steps for geometrical singularities study.
As a next step, safety relief valve of type 1 1/2" G 3" is tested1 . We should point out that
inches are units commonly used in the industry. Both the industrial valve and a transparent
model of it are studied (Fig. 1.9). The valve is modified; the spring is removed and an
apparatus measuring the force is mounted on top of the valve thus, the hydrodynamic
forces applied on the disk at different inlet pressure and lift positions are determined;
the latter was achieved by adjusting several times the disk location. The stresses exerted
by the pressure on a disc of a safety valve are essential for the proper dimensioning of
the spring and the adjusting ring. In transparent facility, flow visualization is feasible,
providing an insight of the flow topology in case of cavitating flow and injected air-water
flow. Additionally, for the case of transparent valve, a closed loop is built, hence the
influence of back pressure is also investigated. The methodology followed for the valve
study is presented in Fig. 1.10.
The reason for initially following the first phase approach is that preliminary goal was
to model the valve as a series of singularities. Since in SRV operation, chocked flow can
either occur IN the nozzle (upstream SRV) or just downstream the valve, as it will be
Denotes 1 1/2 inch inlet flange, 3 inches outlet flange and API “G” orifice effective area (Table E.1 in
1
Appendix E)
11
DN 80 (3″)
Φ 22
DN 40 (1 ½ ″)
Figure 1.9: Plan of transparent modified SRV (1 1/2 G 3 ).

00 00
demonstrated in Chapter 6, expansion and contraction geometries have been chosen for
this study. A key reference for this methodology is the PhD thesis published in German
by B. Shannak who has modeled the safety valve as an orifice and calculated the mixture
loss coefficient of the valve. The author has stated that clear contraction of the flow occurs
in SRV (publications of Shannak (2009) and Shannak et al. (1999)). It is also mentioned
that for sharp orifices two-phase flow in a wide range of air mass flow qualities is not
contracted (between 1.2 and 90 %) while in other limits the flow is contracting which
agrees with remarks and conclusions from Schmidt and Friedel (1997). More information
on his approach and adopted model are given in §2.3.3. The conditions tested during this
project correspond to air mass flow qualities ≤ 1.2 % which leads to conclusion that for
all measurement campaign two-phase flow contraction will occur.
1.4 Overview of the thesis

Prior to presentation, discussion and analysis of experimental results obtained, it is nec-
essary to discuss the basic parameters, assumptions and governing equations concerning
this study. Therefore, at the first part of Chapter 2 the most important quantities and phe-
nomena occurring in two-phase air-water flow are reported. Different pressure drop mod-
els; homogeneous versus separated flow approximation are demonstrated and compared.
Some basic considerations for vapor-liquid flow (since the latter is the most commonly
met situation in SRV operation) are discussed in Appendix F.
Pressure drop under single-phase and two-phase flow conditions in straight and sin-
gular piping systems is of particular interest in this study since the main objective is to
12
1.4 Overview of the thesis 13
Original valve Transparent valve

F Spring removed Pressure drop
Measurements of (static flow behavior)
flowforce on valve disk measurements
Flow visualization
P1=Pa P1=Pb
Flow rate
P0 P0
Flow
Flow
(a) Industrial valve. (b) Transparent valve.
Figure 1.10: Methodology steps for Safety Relief Valve study.
predict pressure drop in SRV in presence of various flow conditions. The core of SRV is
composed by a complex geometry that could possibly be decomposed to simpler geomet-
rical discontinuities. Hence, the behavior of different types of geometrical singularities
mainly in terms of pressure modeling are discussed. In the next part of this chapter, an
analytical description of the different types of relief devices used in industry and their
characteristics is presented. Existing methodologies to predict flow discharge through
SRV is demonstrated and assumptions-limitations of these models are depicted. At the
end of each section, a brief review of existing literature in geometrical singularities and
safety relief valves is carried out.
In Chapter 3, the different experimental facilities (in small and large scale) are demon-
strated; LUCY II and AGATHE for geometrical singularities and LUCY III and AGATHE
II for SRV investigation. Information on the exact test conditions are provided with lim-
itations and application range of tests being pointed out. Measurements performed in
industrial and transparent SRV are discriminated. In the second part of this chapter, mea-
surement techniques used during this study are analyzed; pressure-flow rate measure-
ments and optical probe experiments.
The next two chapters are dedicated in presenting experimental results obtained in
geometrical singularities. They are structured in a similar way; in the first part pressure
evolution curves are plotted and singular pressure drop coefficient is extracted for each
case. Next, flow visualization reveals the flow topology for each case. In the second part,
comparative graphs with literature correlations are given and in the last section a new
methodology for each case is proposed. In the following chapter, outcome from safety
relief valve measurements and simulations are demonstrated.
Chapters 4 and 5 concern expansion and contraction geometries respectively. The
influence of opening angle is discussed (smooth and sudden expansion) both for single-
and two-phase flow. The flow regime in two-phase flow can considerably alter pressure
drop, thus, flow visualization with simple digital and high-speed camera is performed
in order to extract information on the structure of two-phase flow along the pipe. The
results are verified with void fraction profiles attained with optical probe measurements.
Additionally, local velocity and bubble size give information on interaction of the two
phases. Flow pattern maps are deduced for selected configuration in diverse conditions.
In Chapter 6 pressure and force plots are presented for various compressible, incom-
13
pressible and two-phase flow scenarios. Therefore, conclusions on two-phase flow be-
havior of the valve are extracted. The commonly used two-phase methodologies (HEM,
HNE-DS, API520) are applied for two-phase mixture test conditions in transparent and in-
dustrial SRV under investigation and the results are compared with experimental findings.
Flowforce measurements are performed in air, water and two-phase flow mixtures and
results are compared with CFD computations in axisymmetric 2D model of the industrial
valve. Finally, pressure drop correlation for SRV is established function of volumetric
quality of air, liquid flow rate and valve opening.
In last segment of the thesis (Chapter 7) the final conclusions and main achievements
of the thesis are summarized. Additionally, future works and recommendations for even-
tual studies are discussed.
14
2 Theoretical background-Literature
review
In this chapter, a theoretical analysis of two-phase flow in straight pipe flow, flow through
singular geometries and safety relief valves is carried out. The most common correlations
for design calculations involving air-water flow are demonstrated. Various types of geo-
metrical singularities and valve categories are discriminated and the associated theoretical
or empirical models are exemplified. The goal of this chapter is to introduce the common
concepts, definitions, vocabulary used and associated bibliography related to this study.
2.1 Two-phase flows generalities

The main parameters of gas-liquid flows used during this investigation are presented in
the next paragraph (2.1.1). Additionally, the different two-phase flow scenarios in terms
of flow orientation are distinguished and examined.
2.1.1 Parameters describing gas-liquid flows

Due to its fluctuating character, two-phase flow requires the use of several averaging op-
erators (which act on space and time) in order to be described (Delhaye et al. (1980)).
The relevant two-phase quantities, used in the present study, averaged over the pipe cross
section and time interval are defined hereunder.
• Space and time averaged volumetric and mass flow rate

The instantaneous volumetric and mass flow rate Qk and Mk respectively, varying
in time, through a pipe cross section of area A is described by:
Z
Qk (t) = wk dA, (2.1)
Ak
Z
Mk (t) = ρk wk dA, (2.2)
Ak
where k represents the presence of different phase, wk is the axial velocity compo-
nent and ρk the density of phase k.
The time averaged volumetric and mass flow rates (over time interval T) for each
phase k can be expressed by the following equations respectively (Eq. 2.3 and 2.4):
16 2 Theoretical background-Literature review
Z
1
Qk = Qk (t) dt, (2.3)
T T
Z
1
Mk = Mk (t) dt, (2.4)
T T
where over bar symbol indicates time averaging.
• Phase density function

The presence or absence of one phase k at a certain time can be described by the
phase density function. This is a binary function analogous to the intermittency
function used in single-phase flow. The definition of this operator for a specific
position x is the following:


 1 if phase k (for instance the gaseous phase)
Xk (x, t) =ˆ 

is present at point x at time t (2.5)



 0 if phase k is not present at point x at time t.
• Local void fraction and mixture quantities

It is very difficult experimentally to obtain the instantaneous operators over a certain
volume in two-phase flows. Therefore, time-averaging is very often used such as
for the local void fraction of phase k, αlocal which is defined as the average over the
interval T of the phase density function Xk :
Z
1
αlocal (x, k) = Xk (x, t) dt, (2.6)
T T
where symbol T k stands for the time in presence of phase k. The local instantaneous
void fraction requires measurement techniques such as optical probe or electrical
capacitance of a conducting liquid phase to be determined.
Moreover, the cross-sectional void fraction is defined by Eq. 2.7 as the ratio of area
occupied by air over the total area of the pipe. This parameter is simply denoted as
α since it is the most common expression of void fraction.
AG
α= . (2.7)
AG + AL
Volumetric quality β is the parameter that is used to express the relative fraction of
gas to-liquid inside the pipe. This quantity is given as the ratio of the gas volumetric
flow rate to the total volumetric flow rate (Eq. 2.8). For the case of homogeneous
flow (presented in the next sections), it corresponds to an average volume fraction
given from two flow meters (air and water) and should not be distinguished from
cross sectional void fraction α.
QG
β= . (2.8)
QG + QL
16
2.1 Two-phase flows generalities 17
In order to express the true (or mass) quality of a two-phase flow mixture, we must
first define the total mass velocity:
M
G= , (2.9)
A
where M̄ is the time-averaged total mass flow rate.

Thus, the quality or mass quality is given by the following formula:
MG
x=ˆ and M̄ = M̄G + M̄L . (2.10)
M
where x stands for the title and subscripts G and L for the gas and the liquid respec-
tively.
The superficial velocities of the liquid and the gaseous phase respectively, are given
by:
QL Q
JL = , JG = G , (2.11)
A A
where A is the total area occupied by two-phase flow mixture.
• Relation between mixture quantities

Homogeneous flow
Average phase velocities can be expressed function of void fraction and then in
terms of mass quality, mass flow rate and density for each phase:
!
QL G 1−x
UL = = · (2.12)
(1 − α) · A ρL 1 − α
QG G x
UG = = · (2.13)
α · A ρG α
The assumption made in the homogeneous flow model is that liquid and gas-phase
velocities are equal (UL =UG ). Therefore, mass quality and void fraction are related
with Eq. 2.14. This model is also called the zero slip model.
1
α= ρG
(2.14)
1+ 1−x
x
· ρL
Separated flow
With the hypothesis of a slip velocity between the two phases, Eq. 2.14 can be
modified to obtain the void fraction with slip velocity α slip :
17
1
α slip = ρG
(2.15)
1+ 1−x
x
· ρL
·S
where S =UG / UL the slip ratio between the two phases. From Eq. 2.15 it can be
concluded that even for very small values of the mass quality x the void fraction can
reach 50 % for slip ratio=1 while for higher values of S , the void fraction is con-
siderably lower. At the same time, the influence of the pressure can be concluded,
since for lower pressure, thus lower ρG / ρL , the void fraction will be higher for the
same x and S .
A correlation between cross-sectional void faction α and volumetric quality β can
be determined by simply replacing β from Eq. 2.8 into Eq. 2.15:
β
α slip = (2.16)
β + (1 − β) S
Hence, it can be seen that void fraction and volumetric quality are the same only in
the case of homogeneous flow, else the velocity ratio should be known to convert
volumetric quality to cross-sectional void fraction.
The two-phase average velocity of the mixture is deduced by a simple summation
of the velocities of the two phases:
U ∗ = UG + U L . (2.17)
The mixture fluid density is delineated by Eq. 2.18 and Eq. 2.19 for homogeneous
(ρ∗ ) and separated (ρmix ) flow respectively:
ρ∗ = βρG + (1 − β) ρL , (2.18)

ρmix = α slip ρG + 1 − α slip ρL . (2.19)
2.1.2 Description of flow regimes and flow regime maps

A two-phase mixture flowing in a pipe can exhibit several interfacial topologies, such
as bubbles, slugs and films. However, this topology is not always easy to define, which
prevents the flow patterns from being precisely and objectively described.
When dealing with two-phase flows, the flow patterns must be known to model the
physical phenomena as closely as possible. It is obviously impossible to describe for
instance bubbly flows and annular flows with good accuracy by a single model. It is far
better to adopt two different models, each one fitting the individual flow description. Nev-
ertheless, this approach is still difficult because of the transition zone between two flow
patterns and the lack of physical knowledge to describe these buffer zones. In addition to
the random character of each flow configuration, two-phase flows are never fully devel-
oped. In fact, the gas phase expands because of the pressure drop along the pipe. This
18
expansion may lead to a modification of the flow structure, such as an evolution from bub-
bly to slug flow for instance. The flow pattern depends also on the cross section junctions,
restrictions, bends and so on (Delhaye et al. (1980)). As previously mentioned, in this
study, the singularity effect on flow regime will be investigated with flow visualization
and optical probe measurements. However, pressure drop coefficients are only measured
for the same flow regime upstream/downstream the singularity; dispersed bubbly flow.
In two-phase flows, it is very common to use flow regime maps, which are two-
dimensional representations of the flow pattern existence domains. These maps can be
established experimentally, by visualization methods in transparent channels. We have
to point out that flow charts are originally resolved for straight pipes. Thus, for complex
setup, the flow structure is almost impossible to predict. As a consequence, no recom-
mendation can be made concerning the choice of the appropriate flow map for each case,
as no method has been yet proved entirely universal. The coordinate systems used by
the authors in literature are various and so far there is no agreement on the most suitable
coordinate system.
Since the flow patterns depend on the channel orientation, one should study separately
the case of vertical and horizontal flow in the pipe. Thus, different flow regime maps
are established for both cases. The case of vertical upward flow will also be studied
separately from that of downward flow. In present analysis, singularities are investigated
in horizontal direction while safety relief valve is split in two sections; upstream vertical
and downstream horizontal flow. Thus, both cases will be demonstrated since they are
of particular interest for this study; the first case is demonstrated in the next sections
while the second is presented in Appendix G. Additionally, slightly inclined pipe flow is
discussed in Appendix G.3.
Horizontal cocurrent flow

The possible flow configurations that we can meet in horizontal cocurrent flow in a pipe
are more than in the case of vertical flow. This increased number of flow regimes is due
to gravity which plays an important role in horizontal flow. Indeed, the effect of gravity
tends to separate the two phases and create a horizontal stratification.
2.1.2.1 Horizontal flow regimes

The classification of the flow regimes that is shown in Fig. 2.1 was proposed by Alves
(1954). Another classification can be found in Taitel and Dukler (1976b), where plug
and slug flows are in the same category (called intermittent flow). Hence, for horizontal
cocurrent two-phase flow we can distinguish the following situations of the flow:
• Bubbly flow
Bubbles are dispersed in the liquid medium and are moving in the upper part of
the pipe. This is caused by the effect of buoyancy i.e. the difference in the density
between the gas and the liquid. When the liquid velocity becomes very high, the
turbulence intensity disperses the bubbles throughout the tube cross-section. This
flow pattern is called froth flow.
19
• Plug flow
The plug flow regime takes place when increasing the gas flow rate. In this flow
configuration, bubbles coalesce and gas plugs (bubbles of bullet shape) are formed
and move along the upper part of the pipe in a continuous liquid phase.
• Stratified flow
For low liquid and gas flow rate, stratified flow appears with a smooth interface.
The liquid phase flows along the bottom of the pipe and the gas phase above it.
Both phases are continuous.
• Wavy flow
By a further increase of the gas flow rate with the liquid flow rate constant, waves
are created in the pipe and propagate along the interface (wavy flow).
• Slug flow
Slug flow occurs by a further increase of the gas flow rate from the wavy flow. The
waves can then reach the top wall of the tube giving rise to a slug flow. This flow
can be considered intermittent, similarly to the plug regime. However, in this case,
the gas bubbles are larger and the liquid slugs contain many smaller bubbles. The
faster moving gas flow picks up periodic roll waves and forms frothy slugs which
pass through the pipe at a much greater velocity than the average liquid velocity.
• Annular flow
When the gas flow rate is high and the liquid low, annular flow regime can exist.
The liquid flows as a thin film along the inside wall of the pipe while the gas core
at high velocity. More liquid flows on the bottom of the tube than on the top due to
the force of gravity. The liquid surface is not smooth or symmetrical but consists of
wavelets. The gas core must contain dispersed liquid droplets so that the film may
be maintained in the upper part of the pipe by the de-entrainment of this liquid onto
the pipe wall.
• Disperse flow
At high gas velocity nearly all of the liquid is entrained as small droplets in the high
velocity gas phase and the disperse flow regime is formed in the tube.
Images of flow regimes identified in horizontal cocurrent flow obtained from Ghajar
(2004) are illustrated in Fig. 2.2.
2.1.2.2 Flow pattern charts

Baker’s diagram (Baker (1954)) is the most common flow regime map used for horizon-
tal cocurrent flows. The coordinates used by Baker (1954) were modified by Bell et al.
(1970). The modified chart is shown in Fig 2.3. The quantities GG and GL are the superfi-
cial mass velocities of the two phases and are defined by Eq. 2.11 for gas and liquid (with
G and L subscripts respectively).
20
Figure
Figure 2.1: 7.2.
FlowSketches
patterns inofhorizontal
flow regimes fora flow
flow in pipe. of air/water
Adapted mixtures(1983).
by Weisman in a
horizontal, 5.1cm diameter pipe. Adapted from Weisman (1983).
Figure 2.2: Images of flow regimes in horizontal flow in a pipe (taken from Ghajar
(2004)).
The two factors ψ and λ appearing in the abscissa and the ordinate of the map, respec-
tively, depend on the fluid properties and can be defined by the following equations:
!1/2
ρG ρL
λ=ˆ , (2.20)
ρA ρW
!2 1/3
σWA  µL ρW 

ψ=ˆ  . (2.21)
σ µW ρL

21
With the notation ρ, µ and σ, the density, dynamic viscosity and surface tension re-
spectively are represented. Subscripts A and W refer to the physical properties of air and
water respectively at 0.1 MPa pressure and 200C temperature. Therefore, for air-water
flow at 0.1 MPa and 200C, we have λ ≡ 1 and ψ ≡ 1. Values of these parameters for
natural gas and machine oil (medium) mixture are λ '0.8 and ψ '20.7 (ρG =0.862 kg/m3 ,
µG =1.02·10−5 Pa·s, ρL = 940 kg/m3 , µL = 850 Pa·s, σ=32·10−3 N/m). In order to emphasize
to the flow regions that we were mainly dealing with, during this experimental study, in
Fig. 2.3, bubbly and slug flow are highlighted.
Another map was developed by Mandhane et al. (1974) using superficial velocities JL
and JG as coordinates. It is based upon about 6000 experimental data points from pipelines
of diameters between 1.27 cm and 16.51 cm. Flow regimes and maps for slightly inclined
pipe flow are reviewed in Appenix G.
Figure 2.3: Horizontal flow map of Baker modified by Bell et al. (1970).
2.2 Pressure head loss and modeling in straight pipes

The definition of the head loss in a hydraulic system is of major importance. Information
concerning the pressure drop is required for the design/scale up of equipment, both from
the point of view of the fluid dynamics of the system and for the prediction of heat and
mass transfer characteristics. If a system should be designed for a fixed pressure loss, the
relationship between pressure drop and flow velocities establishes the velocity-dependent
parameters. Such parameters are the heat transfer coefficient and mass or heat flux limita-
tions. In the case where the design of fixed flow is required, the head loss determines the
power input required for pumping (Cheremisinoff and Gupta (1983)).
The pressure head loss varies depending on the flow conditions and the geometry of
the system. Elbows, tees and junctions can have a strong influence on the pressure head
22
2.2 Pressure head loss and modeling in straight pipes 23
loss. The orientation of the pipe (horizontal or vertical flow) could also be an important
parameter affecting the pressure drop. Finally, the restriction of the flow can influence the
flow pattern and result in a significant pressure head loss of the system.
Thus, the case of single-phase flow should be separately investigated from two-phase
flow. Additionally, correlations established for calculation of the pressure drop caused by
singularities in the pipe will be presented.
The total head loss in a pipe (for both single and two-phase flow) consists of three
terms; the gravity term, the acceleration term and the frictional term and is, therefore,
given by the following relationship:
! ! !
dP dP dP dP
= + + . (2.22)
dz dz g dz α dz f
Generally, evaluation of the total pressure head loss involves not only the friction
term but also the other two terms. However, since most applications involve long pipe in
horizontal direction (i.e. transport), the term that plays the most important role, both for
single and two-phase flows, is the frictional term. The majority of existing models and
empirical correlations are dedicated to prediction of this term.
Evaluation of frictional pressure drop in single-phase flow for straight pipes has been
well described in literature. Formulations used during this study for the modeling of such
flow upstream and downstream singularities and safety relief valve is given in Appendix
H.
2.2.1 Pressure drop and modeling in two-phase flow

Pressure drop in two-phase gas-liquid flow is an important design parameter in many
engineering applications. Due to its importance, several investigations on the field can be
found in literature. Examples of engineering applications include chemical, nuclear and
petroleum industry, refrigeration and air-conditioning applications.
Calculation of the pressure drop becomes more difficult when dealing with two-phase
flows as equations that govern the physical phenomena are now more complex. The
models that were adopted for prediction of the pressure drop are numerous and the most
important of them will be presented in this section. As previously analyzed, the total
pressure drop for two-phase flow consists of the frictional, acceleration and gravitational
components. It is necessary to know the void fraction to compute the acceleration and
gravitational components whether for the frictional pressure drop, either the two-phase
friction factor or the two-phase frictional multiplier (Lockhart and Martinelli (1949)) must
be known.
Especially, when one considers horizontal two-phase flows in tubes, the hydrostatic
pressure drop (Eq. 2.22) is zero and the pressure drop due to acceleration is negligible.
For the same case but with phase change, the acceleration pressure drop term can be
important.
There are two principal types of frictional pressure drop models in two-phase flow.
The homogeneous and the separated flow model.
23
2.2.1.1 Different types of two-phase flow models

To summarize the various approaches in two-phase flow, we can distinguish (Todreas and
Kazami (1989)):
1. HEM (Homogeneous Equilibrium Model): This model assumes equal velocity for
the two-phases. Additionally, the two phases exist at the same temperature; at the
saturation temperature for the prevailing pressure. The mixture is considered as a
single fluid. The optimum applicability conditions of this model is for high pressure
and flow rates.
2. Thermal equilibrium mixture model with an algebraic relation between the fluid
velocities (slip ratio , 1). The most common model in this category is the drift-flux
model. Is is more useful for low pressure and/or low flow rates under steady- and
quasi-steady-state conditions.
3. Two-fluid (separated) flow model: Allows the two fluids to have both thermal non-
equilibrium and different velocities. Thus, one should solve the three independent
conservation equations for each phase.
2.2.1.2 Homogeneous flow model

This model considers, as explained in §2.1.1, two-phase flow like single-phase flow hav-
ing average fluid properties depending on the mass quality. Thus, the frictional pressure
drop is calculated by assuming a constant friction coefficient between the inlet and outlet
sections of the pipe. It is the simplest approach and its validity depends primarily on the
mixing of gas and liquid; the greatest the phase mixing the highest the accuracy of the
model.
2.2.1.3 Separated flow models

In the case of different flowing velocities of the two phases pressure drop is predicted
with different approach; the separated flow models. Most of the correlations belonging
to this category, are formulated with system variables without any particular reference to
the flow regime and to the physics of the system. A thermodynamic equilibrium between
the two phases is assumed. These models are also called, in contrast to the homogeneous
model, the slip flow models.
Air-water straight pipe flow is herein modeled with Lockhart-Martinelli method and
its validity is checked in the next chapters. Hence, in the next section analytical presen-
tation of this method is following. A more complete description of the other commonly
used two-phase models (for air-water and steam-water flow) is provided in Appendix I.
Lockhart and Martinelli (1949) model This method is the first and most commonly
used method based on the separated flow model. It was proposed by Lockhart and Mar-
tinelli (1949) at Buffalo in New York. It is one of the simplest procedures for calculating
two-phase frictional pressure drop and hold-up. The most important advantage of this
model is that it can be used for all flow patterns. This advantage is at the expense of the
accuracy of the model which is considered to be relatively low.
24
2.2 Pressure head loss and modeling in straight pipes 25
Therefore, Lockhart and Martinelli defined the following variables:

dP/dz dP/dz
ΦG2 =ˆ , Φ2L =ˆ ,
(dP/dz)G (dP/dz)L
ΦG2 (dP/dz)L
X =ˆ 2 =
2
. (2.23)
ΦL (dP/dz)G
All the pressure gradients, in the previous parameters, are frictional pressure gradients,
with (dP/dz)L and (dP/dz)G being evaluated, respectively, for the liquid and the gas phase
flowing alone in the same pipe and with the same mass flow rate (ML and MG ) as in
two-phase flow.
Concerning the flow conditions, four possible flow combinations were considered:
• Turbulent liquid and turbulent gas (tt)
• Laminar liquid and turbulent gas (lt)
• Turbulent liquid and laminar gas (tl)
• Laminar liquid and laminar gas (ll)
Using the Blasius (1913) law for turbulent flows, we obtain:

!0.25 2
µL
!
dP GL
= 0.3164 ,
dz L GL D 2ρL
!0.25 2
µG
!
dP GG
= 0.3164 .
dz G GG D 2ρG
The following formula can therefore be extracted:
!0.25 !1.75
µL 1−x ρG
Xtt2 = .
µG x ρL
The authors plotted φGtt and φLtt for turbulent flows, as well as the void fraction in
function of Xtt . In a similar way, the other three possible flow combinations were also
considered (lt, tl, ll).
These curves could be approximated by:
C 1
φ2L = 1 + + 2, φG2 = 1 + CX + X 2 .
X X
And for the void fraction:
X
1−α= √ ,
X 2 + 20X + 1
where the values for C are:
1. Turbulent liquid and turbulent gas C=20
2. Laminar liquid and turbulent gas C=12
25
3. Turbulent liquid and laminar gas C=10

4. Laminar liquid and laminar gas C=5
The results obtained by Lockhart and Martinelli were based on data of horizontal flow
in adiabatic two-component systems at low pressure. A graphical representation of the
model is shown in Fig. 2.4.
Figure 2.4: Two-phase flow multiplier Φ2L function of the Lockhart and Martinelli (1949)
parameter.
2.3 Flow in geometrical singularities

2.3.1 Description of different types of singularities
The term singularities is used to describe elements of a pipework system in which there
is a singular change from straight channels of uniform cross-section. Such singulari-
ties include sudden or smooth contractions or expansions, bends, partially opened valves
and throttling and/or metering devices such as nozzles, venturies and orifices (Schlünder
(1983)).
The singularities can be split in the following categories:
1. Singularities that interrupt the flow through a local change of cross-section which
otherwise is of straight piping such as bends and open valves or singularities placed
in the channel with the purpose of measuring flow quantities; examples are orifice
plates and venturies. In order to establish the pressure change for the latter case,
pressure at the maximum velocity downstream should be determined (vena con-
tracta position). For this type of singularities, only irreversible pressure head loss
should be resolved (no change of section).
26
2.3 Flow in geometrical singularities 27
2. Singularities leading to a change in cross-section. Examples are sudden and smooth

contraction and expansion; both irreversible and reversible pressure change take
place.
There is a variety of models and correlations involving two-phase flow in singularities

existing in literature. However, for most cases, when most specific informations on the
type and nature of the flow are not provided, the use of homogeneous flow assumption is
advised.
Thus, for single-phase flow, the pressure change is given by the formula:
1
∆PS P,L = k ρL U 2 , (2.24)
2
where k is the pressure loss coefficient and U the average velocity of the fluid.
If we substitute the mixture density (given by Eq.2.18) into Eq.2.24, the two-phase
flow pressure drop through the singularity will be given by:
1
∆PT P = k ρ∗ U 2
2
1
= k ρG β + (1 − β) ρL U 2

2
1
= k ρG βU 2 + ∆PS P,L (1 − β)
2 (2.25)
ρG
" ! #
Eq.2.24
= ∆PS P,L · β −1 +1
ρL
| {z }
Homogeneous multiplier
As it is indicated, second term of this formula represents the homogeneous multiplier.
2.3.2 Modeling of pressure change in singularities

Singular pressure change in single-phase flow has been extensively studied. The work of
Idel’Cik (1986) is a commonly used practice and a reference study in this topic. More
details and examples about single-phase flow in singularities are given in Appendix J. In
this paragraph the more complex scenario of two-phase in geometrical discontinuity is
analyzed.
The most common types of singular geometries met in the industry and academic
studies are presented. These geometries are sudden and smooth contraction and expansion
(Fig. 2.5).
2.3.2.1 Sudden enlargement

A sudden enlargement of the tube results in a pressure recovery step. The theoretical
formulas of the pressure change in sudden enlargement were derived from Collier (1972)
and Hewitt and Hall-Taylor (1970). If an average pressure along the cross-section is
27
Smooth divergence 1 2 Smooth convergence
P1 P2
A1 A2
α
Sudden expansion Sudden contraction
P1 P2
A1 A2
Figure 2.5: Different types of singular geometries.
assumed (P1 upstream and P2 downstream), as shown in Fig. 2.5, the total pressure change
will be given by Schlünder (1983):
∆PT P = ∆PIT P + |
|{z} ∆P RT P
{z } (2.26)
Irreversible Reversible
where ∆PIT P the irreversible (due to friction) pressure change and ∆PRT P the reversible
pressure change (due to acceleration).
For homogeneous flow and if momentum balances are applied upstream and down-
stream the singularity, assuming that pressure is constant along a cross-section, we obtain:
" !#
1 2 vG
∆PT P = G1 (1 − σ) vL 1 + x
2
−1 (2.27)
2 vL
The two terms of Eq.2.26 are presented as:
" !#
vG
∆PIT P = −G21 σ (1 − σ) vL 1 + x −1 (2.28)
vL
and
1 2 " vG
!#
∆PRT P = G1 2σ (1 − σ) 1 − σ vL 1 + x
2
−1 (2.29)
2 vL
where σ=A1 /A2 the surface area ratio, G1 the mass flux upstream, vG and vL the
specific volume of gas and liquid respectively.
If the separated flow model is applied (for a constant value of the void fraction):
(1 − x)2 vG x2
" #
P1 − P2 = −G21 σ (1 − σ) vL + (2.30)
1−β vL β
with ∆PIT P and ∆PRT P formulated as:
28
n (1 − x)2 vL x2 vG
∆PIT P = − G21 (1 − σ) +
1−β β
h i
(1 − σ) x3 vG2 /β2 + (1 − x)3 v2L / 1 − β2 o (2.31)
−
2 [(1 − x) vL + xvG ]
h i
G21 1 − σ2 x3 vG2 /β2 + (1 − x)3 v2L / (1 − β)2
∆PRT P =
2 [(1 − x) vL + xvG ] (2.32)
The homogeneous flow assumption tends to overpredict the static pressure recovery
while the separated flow model provides more accurate results. As already mentioned the
two models assume a constant void fraction value along the expansion which, in many
cases, can be a false hypothesis.
The previous equations can be rewritten in the following final form for total pressure
change:
Jannsen and Kervinen (1966):
G21 ρL
" !#
∆Ptot =− (1 − σ) 1 + x
2
−1 , (2.33)
2ρL ρG
where G1 the mass flux upstream the singularity, ρL the density of water, ρG the density
of air, σ the area ratio and x the mass quality of air.
The prediction of singular pressure change across axisymmetric sudden expansion
geometry can also be obtained from the following model:
Chisholm (1969):
G21
!
C 1
∆P st = − σ (1 − σ) (1 − x) 1 + + 2 ,
2
(2.34)
2ρL X X
where
1 − x 2 ρG
!
2
X , ,
x ρL
!0.5  
  ρL 0.5
 !0.5 
ρ ρ ρ
!
L − G G
C = 1 + 0.5 +  .
 
ρL ρG ρL 
 
Both models rely on the assumption of a homogeneous flow. Chisholm’s model

(Chisholm (1969)) overestimates the pressure change while the formula proposed by
Jannsen and Kervinen (1966) gives more realistic results. This was reported by Velasco
(1975).
29
1 C 2
Flow
P1 P2
A1 A2
Sudden contraction
Figure 2.6: Sudden contraction geometry
2.3.2.2 Sudden contraction
When a sudden contraction of the pipe occurs a static pressure drop step will take place.
A contraction of the flow occurs at the indicated position (C) and the so-called vena
contracta is formed (CC ). The flow in a sudden contraction is illustrated in Fig. 2.6.
The correlation for sudden convergence, as described from Jannsen and Kervinen
(1966), is recalled:
!2
G2 ρL
 " !#
 1 1 
∆PT P = 2 − 1 + 1 − 2  1 + x . (2.35)
σ ρG

2ρL CC
where Cc is the contraction coefficient defined as Cc =Ac /A1 where Ac the flow area in
the vena contracta. A typical value of this parameter is equal to 0.64.
The classical correlation for estimating the contraction coefficient Cc was given by
Weisbach (1845):
!3
A2
Cc = 0.63 + 0.37 (2.36)
A1
2.3.2.3 Smooth change of cross-section
When a smooth change of cross-section occurs, the pressure change can be calculated by
the normal momentum equation for the frictional loss with an additional term that takes
into account the acceleration due to the change of section. Therefore, if no flow separation
takes place (i.e. diffuser 5-7 ◦ ) and for an homogeneous assumption in steady-state, one
can derive (Schlünder (1983)):
dP τ0 P G2 dA
!
2 d 1
− = +G − + gρ∗ sinγ (2.37)
dz A dz ρ∗ ∗
Aρ dz
where A the cross sectional area, τ0 the wall shear stress, G the total mass flux, ρ∗ the
homogeneous flow density (given by Eq. 2.18), γ the opening angle of the cone and g the
gravitational acceleration.
The separated flow model for a constant wall shear stress leads to the following equa-
tion:
30
dP τ0 P (1 − x)2
" 2 #
2 d x
− = +G +
dz A dz ρG α ρL (1 − α)
G 2 x2 (1 − x)2 dA
" #
(2.38)
− + + gρmix sinγ
A ρG α ρL (1 − α) dz
where x the mass quality of the gas, α the local void fraction and ρmix the separated
flow mixture density given by Eq. 2.19.
In Appendix J.2, the formulas to calculate pressure change in singularities with change
of cross-section are derived.
2.3.3 Literature survey on geometrical singularities

Although several studies have been performed in multiphase in geometrical singularities,
since the phenomena occurring are very complex, published data, models adopted and
physical analysis still don’t appear adequate. The presence of more than one phase from
one hand and from the other hand, the change of section or geometrical restriction will
enhance the difficulties of studying this subject.
In most cases the main interest is focused in determining the pressure difference cre-
ated in single and two-phase flow due to the singularity. Therefore, pressure drop corre-
lations are established. An important deviation from experimental findings (which can be
of the order of 50 % and more) is often observed. Hence, since no general correlations for
all types of singularities are yet found (large application range), these correlations should
be used with care.
An important amount of publications is dedicated in determining the local void frac-
tion, since knowing this quantity is often necessary in pressure drop correlations. The
most common technique used in these papers is optical probe measurements. If, addition-
ally, the local bubble velocity has to be established a dual optical probe is used. When
the purpose of the study is obtain the slip ratio, this technique is combined with hot film
anemometry (to determine liquid velocity). Knowledge of the bubble size diameter and
velocity can be of major importance for building a database that can be used in numerical
simulations and modeling.
The aforementioned investigations can be divided in the following two categories;
expansion and contraction geometries.
2.3.3.1 Expansion geometries

Schmidt and Friedel (1996) have studied sudden expansions in duct areas. They have
established a new model to calculate the two-phase pressure change caused by the singu-
larity. The model is tested using air-water, aqueous glycerol, watery calcium nitrate and
refrigerant R-12 experimental data. The new model is found to be accurate enough to
be used in engineering applications. Aloui and Souhar (1996), Aloui et al. (1999) have
carried out experiments in axisymmetric sudden expansion geometries and determined
the void fraction, bubble size and velocities. Flow visualization has proved that bubbly
31
flow changes from a asymmetric configuration to a symmetric one above certain values of
the volumetric quality. Ahmed et al. (2007, 2008) have focused their interest in the flow
structure and pressure losses in sudden expansions. The working fluids were air and oil
and the area ratios σ=0.0625 and 0.25. They have concluded that the phase redistribution
immediately downstream of the expansion and the developing length are strongly depen-
dent on the upstream flow pattern and the sudden expansion area ratio. Additionally, the
pressure recovery was found to be dependent both on the wall shear stress and the wall
pressure in the developing region immediately downstream the expansion.
Lottes (1960) in his paper has presented in detail a few methods that are available in
literature to evaluate the expansion losses in two-phase flow. He has stated that the regions
of the highest loss are the regions of the highest void fraction and liquid velocity since the
loss is proportional to the square of the liquid velocities.
2.3.3.2 Contraction geometries

Fossa and Guglielmini (1998) focused their study in void fraction measurements and
phase distribution for horizontal sudden contraction. An air-water mixture was investi-
gated with pipe inner diameters 70/60 mm and 70/36 mm. They have concluded that the
characteristics of the flow restriction deeply modify the flow structure upstream and down-
stream the discontinuity. They have studied thin and thick orifices as well for s=0.73 and
0.54 and remarked an increase of the void fraction by almost 50 % compared to straight
pipe values.
Bertola (2002, 2004) has conducted experiments in horizontal abrupt area contraction
with diameters of 80 and 60 mm for upstream-downstream sections respectively using a
single-fiber optical probe. The water flow rate tested was 3 kg/s and the volumetric qual-
ities of air varied between 0.2 and 0.8. These measurements have allowed to acquire a
cross sectional average void fraction by numerical integration of the local values. A con-
siderable change in the distribution of the two phases is observed due to the convergence
geometry.
Schmidt and Friedel (1997) have studied sudden contractions in duct areas. They have
proposed a model to predict the pressure drop in contractions including all the relevant
physical properties. This correlation, in contrast to other methods existing in literature,
takes into account the entrainment of liquid in the gas stream. Guglielmini et al. (1997)
have reviewed the horizontal and vertical correlations in sudden contraction singularities.
Particular emphasis is given to a peculiar behavior encountered at medium-high area ratios
and medium-low mass flow rates. Local void fraction measurements allowed to detect the
influence of the contraction on the structure of the flow along the duct.
2.4 Industrial valves

2.4.1 Description of different types of valves
Valves are hydraulic components very often used in the industry in circuits containing
tubes, reservoirs and other devices to control (completely open, partially obstruct or stop)
the flow rate of gas or liquid flow. Examples of different types of valves are shown in Fig.
2.7.
32
2.4 Industrial valves 33
Ball valve Safety valve
Globe valve Gate valve Butterfly valve
Figure 2.7: Different types of valves.
The model of Chisholm (1971) predicts the two-phase multiplier for several types of
valves:
ρL
!h i
ΦL0 = 1 +
2
− 1 Bx (1 − x) + x2 , (2.39)
ρG
and the proposed values for the parameter B are:
• B=0.5 for thin orifices and control valves
• B=1.5 for thick orifices, butterfly, plug, diaphragm and safety valves
• B=2.3 for ball valves
This study is concentrated in testing and modeling safety relief valve.
2.4.2 Introduction to Pressure Relief Valves (PRV)

The primary purpose of a Pressure Relief Valve (named PRV from now on) is the protec-
tion of life and property by venting process fluid from an overpressurized vessel or adding
fluid (air) to prevent formation of a vacuum strong enough to cause a vessel to collapse.
The proper sizing, manufacture, assembly, testing, installation and maintenance of a PRV
are critical for optimal protection of a vessel or system (Tyco (2008)).
The nomenclature used to categorize PRV is of this type: a" Ab" where a and b denote
the inlet and outlet flange in inches respectively and A the orifice effective area1 . The
1
Table E.1 in Appendix E
33
valve used in this research project is a Weir 1 1/2" G 3" API spring loaded Safety Relief
Valve (SRV). This valve meets with the requirements of ASME section VIII valve design;
it provides requirements applicable to the design, fabrication, inspection, testing, and
certification of pressure vessels operating at either internal or external pressures exceeding
15 psig (ASME VIII-Div.1 (2010)). The different components of Safety Relied Valve are
presented in Fig. 2.8(a). The main components of the valve are; the spring that keeps the
valve closed until the maximum allowable pressure in the protected system is reached, the
valve disk and the nozzle that drives the flow to the main valve body.
Principal components of SRV Spring

Valve disk
1. Body 9. Spindle head A2
2. Bonnet 10. Spindle
3. Cap 11. Guide
4. Nozzle 12. Adjusting screw Huddling
5. Adjusting ring 13. Nut chamber
6. Adjusting ring pin 14. Gasket
7. Disc 15. Spring washer A1
Adjustment ring
8. Disc holder 16. Spring Nozzle
(a) Different components of SRV (AFIR- (b) Detailed view of nozzle-valve disk
CETIM (1999)). components of SRV.
Figure 2.8: Conventional spring loaded SRV.
Working principle of SRV

The simple schematic presented in Fig. 2.8(b) demonstrates the working principle of a
SRV. The valve disk is constantly pressed against a nozzle by a spring. When the pressure
forces on the upstream facade of the disc are inferior to the force applied by the spring,
the valve remains closed. If the pressure in the system under protection prevails over a
predetermined level, the force acting on the valve disk becomes higher than the elastic
forces of the spring and the valve opens. Thus, the pressure of the system to be protected
is reduced to an acceptable value.
34
The surface of the valve disk A2 is designed to be larger from that of the nozzle A1 .
The fluid entering from the nozzle to a larger area, for a constant system pressure, will
result in a higher force that will preponderate for the spring force keeping the disk closed.
Therefore, a fast opening of the valve will take place. Additionally, momentum influence
caused by the change in flow direction will further enhance the valve lift (bend effect due
to centrifugal forces). As a result of the larger area, the valve will not close until the
system pressure drops to a certain percentage of the set pressure; the so-called blowdown
pressure (Tyco (2008)).
The annular pressure chamber between the nozzle exit and the disk or disk holder is
called huddling chamber as demonstrated in Fig. 2.8(b). This control chamber geometry
will determine when the valve will close. Thus, according to the desired operating con-
ditions, most valves are equipped with an adjustment ring. This ring is assembled to the
nozzle or guide of a direct spring-loaded SRV used to control the opening characteristics
or the reseat pressure of the valve. The valve ring controls the huddling chamber geometry
which as a consequence will vary the valve opening and the blowdown pressure. Hence,
the desired lifting force is achieved resulting in pop-action of the valve, mainly required
in compressible flow operation. If the design maximizes lift flowforce, then blowdown
will be long. If the design objective is to minimize blowdown, then the disk effort will be
diminished.
2.4.3 Relief Devices (RD)-detailed analysis

In this section, a classification of RD and an overview of their most important features
is provided. In Table 2.1 a review of the types of RD in a tree form is presented. Relief
devices are mainly distinguished in re-closing pressure relief valves split in safety valves,
relief valves and safety relief valves depending on the fluid used and non-reclosing devices
such as rupture disks. A thorough analysis of the types of relief devices encountered in
the industry can be found in Tyco (2008).
2.4.3.1 Safety Relief Valve (SRV) important parameters

The present study will be concentrated in the use of Safety Relief Valves (SRV) since both
compressible and incompressible fluids are investigated. Additionally, the two-phase flow
behavior will be examined in detail.
The definition of the main quantities characterizing a SRV are herein sorted (Tyco
(2008)):
• Opening pressure is the value of the increasing static pressure of a SRV at which
a lift can be measured or a continuous discharge can be seen, heard or felt.
• Popping pressure is the value of the increasing static pressure of a SRV at which
the disk opens at a faster rate compared to the matching movement at higher or
lower pressures.
• Start-to-leak or simply leak pressure is the value of the increasing static pressure
of a SRV at which the first bubble appears when the valve is tested in air and the
outlet is sealed with water.
35
• Set pressure P set is the value of increasing inlet static pressure at which a SRV
displays the opening of the valve i.e. opening pressure, popping pressure, start-to-
leak pressure etc.
• Overpressure is the pressure increase over the set pressure of a PRV, usually ex-
pressed in a percentage of the set pressure.
• Closing or reseating pressure Pc is the value of decreasing inlet static pressure for
which the valve disk re-establishes contact with the seat and the lift becomes zero
again.
• Superimposed back pressure is the static pressure existing at the outlet of a PRV
at the time it is required to operate. It results from the pressure existing in the
downstream system and it can be constant or variable.
• Built-up back pressure Pb is the pressure existing at the outlet of the PRV caused
by the flow through the valve itself into the discharge system.
• Back pressure is the static pressure existing at the outlet of a PRV due to the pres-
sure in the discharge system. It consists the sum of the superimposed back pressure
and the built-up back pressure.
• Blowdown P set -Pc is the difference between the actual set pressure of a PRV and
the actual reseating pressure, usually expressed as a percentage of the set pressure.
• Chatter is the abnormal, rapid motion of the moving parts of a PRV in which the
valve disk contacts the seat.
• Relieving pressure is the summation of the set pressure and overpressure.
• Discharge coefficient Kd is the ratio of the measured discharged flow rate from the
PRV to the theoretical discharged flow rate from a nozzle of equivalent dimensions
in the same flow conditions.
The correct sizing of a SRV is a very important issue in the industry. Undersized
valves will lead to lower than required discharged fluid and therefore overpressure in the
system or process under protection which will result in a possible breakdown of the equip-
ment. On the contrary, oversized SRV will cause a higher than predicted mass flow rate
in the downstream section and since the downstream piping will be undersized, overflow
can occur. Thus, a thorough and precise research on this topic is essential.
At this point, the use of two other commonly used types of valves; the pilot operated
valves and balanced bellows valves is demonstrated.
36
Types of Relief devices (RD)
Reclosing pressure relieving devices Non-reclosing pressure

Pressure Relief Valves (PRV) relieving devices
2.4 Industrial valves
Types of PRV Safety Valve (SV) Relief Valve (RV) Safety Relief Valve (SRV) • Rupture disk
• Breaking Pin
• Buckling Pin
• Shear Pin
• Fusible Plug
• Low lift PRV Rapid opening Gradual opening or Rapid opening or closing • Frangible disk
• Full lift PRV or closing closing (incompressible or gradual opening or • Bursting disk
• Reduced bore PRV (compressible fluids) closing (incompressible • Direct spring-loaded
• Full bore PRV fluids) or compressible fluids) • Pilot operated
• Direct spring-loaded PRV
• Pilot operated PRV
• Conventional direct spring-loaded PRV Gas Liquid
• Balanced direct spring-loaded PRV
Gas or Liquid
• Internal spring PRV Low lift RD
• Temperature and pressure RV Full lift RD
• Power actuated PRV Reduced bore RD
Full bore RD
Table 2.1: Classification of relief devices Tyco (2008).
37
37
2.4.3.2 Pilot Operated Relief Valve (PORV)

The Pilot Operated Relief Valves (PORVs) consist of a main valve with a diaphragm or
piston or bellows operated seat and a pilot. These valves can also be configured for very
low set pressures using a weight loaded arrangement (Fig. 2.9(a)). Typically,the valve
would be equipped with a pilot for pressure relief and weight loaded for vacuum relief or
with two pilots, one for each relief function. Under normal conditions, the pilot allows
the system pressure or vacuum into the diaphragm chamber. Since the diaphragm area is
greater than the seat plate area, the plate is held closed. When the set pressure or vacuum
is reached, the pilot actuates to vent the diaphragm chamber. The pilot, upon actuating
at set point, must be able to reduce the diaphragm chamber pressure enough to allow
inlet pressure times seat area to exceed the closing force equal to the reduced diaphragm
chamber pressure times effective diaphragm area, allowing the valve to open. This causes
the seat plate to lift.
The pilot operated pressure/vacuum relief valve has several advantages. As the system
Tyco Pressure Relief Valve
pressure Engineering
increases, Handbook
the force holding the seat plate in the closed position increases. This
Chapter 3 - Design Fundamentals
allows the system operating pressure to be increased without danger of increased seat
leakage in the main valve (Tyco (2008)). In Fig. 2.9(b) an example of a PORV installed
in a compressed air reservoir is shown.
alves
Pilot
ty device
m during an Diaphragm
e events chamber
ressure or
ease beyond its
operating at all
failure when
e source of Diaphragm area
ve is the
Seat area
Vacuum
ressure or
ease to a Pressure
may be the only
failure. Since Main Valve
lexity
n of the Type
(a) Design of 9300Tyco (2008)
a PORV (b) Picture of a PORV installed in vessel
ble. The Figure 3-14
ermined set into the diaphragm chamber. Since the diaphragm
at a specified Figure 2.9: Pilotarea
operated valve (PORV).
is greater than the seat plate area, the plate is held
when the closed. When the set pressure or vacuum is reached,
o a safe level. the pilot actuates to vent the diaphragm chamber. The
nsistently pilot, upon actuating at set point, must be able to reduce
2.4.3.3
the diaphragm Balanced Bellows
chamber pressure Valve
enough to allow inlet
pressure times seat area to exceed the closing force
equal
Thetoeffect
the reduced
of thediaphragm chamberback
superimposed pressure times (back pressure in the downstream section
pressure
essure/vacuum
effective diaphragm area, allowing the valve to open. This
pplications when the valve is closed) is very important in SRV and it is highly recommended to
causes the seat plate to lift.
systems
or in ASME
take into consideration this parameter. The back pressure can result from the presence
The pilot operated pressure/vacuum relief valve has several
me models of a pressurized
advantages. system
As the system connected
pressure increases,downstream of the
valve or other RV connected to
the force
API 2000 and holding the seat plate in the closed position increases.
the primary valve. In the case of a conventional valve (without bellows), the influence
nternational This allows the system operating pressure to be increased
ulations. These
of a danger
without fixed back pressure
of increased can beinequalized
seat leakage by reducing
the spring force. Therefore, the
the main valve.
as well as A unique combination of pressure actuated diaphragms
38 fixed and variable orifices provide the pilot with
using
relief valves either pop or modulating action.
ure or vacuum In modulating pilot valves, minor overpressure conditions
re and vacuum are controlled without fully opening the main valve. This
limits fluid loss and system shock. Another advantage
of pilot operated pressure/vacuum relief valves is the
reduced cost of larger valve sizes. The large spring and
acuum pilot associated envelope is replaced by a small pilot, thus
diaphragm or reducing the mass and cost of the valve.
Back pressure, which may occur after the valve is open the valve’s performance. As a res
and flowing,
2.4 Industrial valves is called dynamic or built-up back pressure. is generally39
not a concern for this
This type of back pressure is caused by fluid flowing from
the pressure relief valve through the downstream piping
back pressuresystem.
plus theBuilt-up
spring back
forcepressure will not
will balance theaffect
inlet the
set valve
pressure. However, when
opening pressure, but may have an effect
the superimposed back pressure is variable, a so-called balanced on valve lift bellows
and SRV shall be
flow. On applications of 10% overpressure, balanced
used. In Fig. 2.10 examples of balanced bellows and balanced piston RV are depicted. A
bellows or balanced piston designs are recommended
bonnet exists in bothbuilt-up
when cases toback
assure that the
pressure is bellows
expected ortopiston
exceed are10%
exposed in atmospheric
pressure. Hence, if a leakage from the bellows or
of the cold differential test pressure (CDTP). piston occurs, this will be observed
from the bonnet. The presence of the bonnet or piston will eliminate the effect of varying
In addition to offsetting the effects of variable back
back pressurepressure,
on the setthepressure. Nevertheless,
bellows or piston acts toa seal
backprocess
pressure due to the flow from
fluid
downstream section when thetovalve
from escaping is openand
atmosphere can isolates
occur. The
the aforementioned
spring, pressure will
only influencebonnet
the liftand
andguiding
upstream flow but
surfaces notcontacting
from the set pressure.
the process
fluid. This is especially important for corrosive services.
Figure 3-
Closed discharge systems (freq
outlet silencers are utilized) pre
circumstance relative to allowab
closed type system is employed
pressure should generally not b
Bonnet 10% of the valve’s inlet pressure
vent maximum of 15% may be accep
open All built-up back pressures exce
inlet pressure in a closed discha
reviewed and approved by Cros
Bellows
Crosby Style JBS-E Crosby Series BP

Figure 2.10: Balanced bellows (left) and balanced piston RV (right) Tyco (2008).
Balanced Pressure Relief Valves
Figure 3-17
Tec
2.4.4 Single-phase flow aspect of SRV Copyright © 2008 Tyco Flow Control. All rights reserved.
Investigation of SRV consists mainly on determining a crucial parameter for the design
of the valve, the discharge coefficient Kd , since this will allow properly sizing the valve.
The formula for calculating this parameter is given in Eq. 2.40:
n 0
P Qm
Qm
i=1
Kd = (2.40)
n
0
where n the number of experiments and Qm the experimental discharged flow rate and
Qm the theoretical mass flow rate.
39
Two different cases shall be distinguished; incompressible and compressible fluids.

Additionally, due to the restriction of the flow, locally chocked flow can occur (more
details on this phenomenon are given in §2.4.6). Therefore, the sub-critical and critical
flow conditions should be discriminated. The definitions provided in the next paragraphs
are in agreement with the standard ISO 4126-1 (2004).
According to the standard the certified discharge coefficient after exhaustion Kdr should
not be higher than 90 % of the discharge coefficient determined from the measurements.
Therefore:
Kdr ≤ 0.9Kd . (2.41)
!(k/(k−1))
Pb 2
Critical: ≤ (2.42)
P set k+1
!(k/(k−1))
Pb 2
Sub-critical: > (2.43)
P set k+1
where the ratio of specific heats k= CCVP . With C the specific heat capacity of a gas is
noted, suffix P and V refer to constant pressure and constant volume conditions respec-
tively. The critical value for air at t=20 ◦ C is 0.528.
2.4.4.1 Incompressible fluid-Liquid

After applying the Bernoulli equation, the theoretical mass flow rate passing through an
ideal nozzle for incompressible fluid can be derived:
Qm = A 2ρL (P set − Pb )
p
(2.44)
The previous formula is valid using SI units. If commonly used units are applied, the
formula becomes:
Qm = 1.61A ρL (P set − Pb )
p
(2.45)
with Qm is the theoretical mass flow rate in kg/h, A the nominal section of the flow in
mm2 , P set the generating pressure in bars, Pb the back pressure in bars and ρL the density
of water in kg/m3 . With factor 1.61 being calculated with conversion from SI units as:
√
3600 2
√ .
10 105
2.4.4.2 Compressible fluid-Gas

The theoretical mass flow rate for compressible fluid at critical flow conditions is:
r
M
Qm = P setC = 0.2883C ρG P set
p
(2.46)
ZT 0
and
40
s
!(k+1)/(k−1)
2
C = 3.948 k (2.47)
k+1
where Qm in kg/h, P set and Pb in bars, A the cross-section in mm2 , M the molar mass in
kgkmol−1 , T0 the temperature in ◦ K and Z the compressibility factor in the real pressure
and temperature conditions of valve opening. One should stress attention to a careful use
of these units to achieve the correct result.
For the sub-critical conditions Qm is:
r
M
Qm = P setCKb = 0.2883CKb ρG P set
p
(2.48)
ZT 0
where
v
u
u 2/k (k+1)/k
Pb
− PPsetb
u
u
u 2k
t k−1 P set
Kb = (k+1)/(k−1) (2.49)
2
k k+1
with Kb denoting the back pressure correction factor. This parameter is used only in
subsonic flow which occurs when the back pressure ratio (nb =Pb /P set ) is higher than the
critical pressure ratio (ncrit =Pcrit /P set ).
2.4.5 Cavitation
Frequently, in high velocity liquid flows, the pressure can locally fall sufficiently low so
that bounded vapor bubbles or gas is formed; this phenomenon is called cavitation. In
order to characterize how close pressure in liquid flow is to vapor pressure, the so-called
cavitation number can be defined:
2 · (PL − P sat )
σ0 = (2.50)
ρL U L2
where PL the pressure of the liquid, ρL the density of the liquid and P sat , UL the satu-
ration pressure and the velocity of the liquid respectively.
A simple physical explanation of cavitation is given in Fig. 2.11 for the case of water.
The diagram of the triple point shows that evaporation occurs either by increasing the tem-
perature for the same pressure (boiling) or by a pressure drop for a constant temperature.
The lateral displacement on the graph is the case of cavitation.
The symptoms of cavitation are usually increased noise emission, valve and pipe com-
ponent erosion or low-frequency mechanical vibration in the valve and the connected
pipeline. For instance, cavitation can occur in valves, blades of turbines, propellers of
ships and other applications. Under these conditions, in particular, neglecting details can
result in negative influences on plant performance and costs of ownership. The presence
of cavitation is usually identified by bubbles that are locally formed and then collapse.
The material surface, depending on its structure, is deformed, loosened and eventually
eroded in particles in various ways due to the frequent strain from the pressure waves
created by the micro-jet occurring when the bubble collapses.
41
water
vapor pressure p
critical point
liquid Boiling
solid Cavitation
vapor
temperature T
triple point
Figure 2.11: Critical point diagram of water. Cavitation and boiling phenomena.
Example of corrosion and cavitation bubbles in SRV observed during this study are
shown in Fig. 2.12(a) and Fig. 2.12(b) respectively. Appearance of rust in control valve
and valve plate is demonstrated in Fig. 2.12(c). Finally, vapor bubbles formation in
globe valve is exemplified in Fig. 2.12(d). More instances of cavitation can be found in
Oertel et al. (2009). In the present investigation, cavitation has been observed for most
measurement points in SRV. This phenomenon has not been measured or modeled with
any particular physical interpretation. However, its presence is included in ω methodology
and hence is taken into account for discharged mass flux calculations §2.4.7.
2.4.6 Critical flows

Critical flow can easily be reached not only in compressible flows but also in two-phase
flow installations. Even a small pressure difference occurring in a flow restriction (such
an orifice or a valve) can cause a sonic blockage. The calculation of critical flow rates,
the so-called chocking flows and understanding of the phenomena related to wave propa-
gation are important for the design of equipment involving two-phase flows. In practice,
critical flows are often associated with a phase change in steady-state flow such as in re-
frigerating and desalination plants due to flashing phenomena. The incidence of critical
flows in the safety analysis of nuclear reactors is of major importance; critical flow rate
calculations are performed while the study of the loss of coolant accident-LOCA (Delhaye
et al. (1980)).
The speed of sound is defined as (Joukowsky (1898)):
s s
∂P K
c= (adiabatic) or c = Korteweg (1878) formula, (2.51)
∂ρ ρ
where c and K the fluid bulk modulus of elasticity (compressibility factor). Thus, the
speed of sound increases with the stiffness of the material, and decreases with the density.
42
At 2mm, the DP point of view (Figure 5) show that the curves remain nearly linear, no
significant Cavitation
influence due to cavitation can be observed. From the Q² point of view
instead (Figure 6), since choking impedes the flow rate, low downstream pressure (i.e.
high cavitation level) curves are shifted to the left. The effects of cavitation are thus seen
as an increase of the transverse force. In other words, to overcome the additional
pressure drop created by cavitation and to sustain the flow rate, an increase in the
Cavitation
Cavitation bubbles bubbles Cavitation bubbles
upstream pressure (hence DP) is required, consequently the transverse force increases.
At 16mm, the DP point of view illustrates (Figure 9) that as cavitation appears with
increasing DP the curves become nonlinear and cavitation reduces the transverse force.
From the Q² point of view instead (Figure 10), cavitation has no effect on the transverse
force. In other words, the additional pressure drop created by cavitation reduces the
pressure drop felt by the stem and thus the transverse force.
At 6mm (Figure 7 and Figure 8), it exhibits a transitional behavior between the two
Cavitation
previous observed openings. Other graphs at 3 and 4mm opening (not shown in this
paper) confirm the transitional behavior.
Based
(a) Corrosion on the
in valve experimental
disk of SRV duedata, the effect(b)
to cavita- of Vapor
cavitation on the
bubbles transverse
formed due toforce depends
cavitation in SRV
tion on the valve disc position.
Cavitation Erosion Damage to

Case-hardened Valve Plate
Figure 11. cavitation pattern at 2mm opening under high Figure 12. cavitation pattern 16mm opening under high
(c) Appearance
flow rateof corrosion due to cavitation in con- (d) rate
flow Cavitation in globe valve (from Ferrari
trol valve-taken from Stares (2007) (left) and in case- and Leutwyler (2008))
hardened valve plate (right)
Figure 2.12: Examples of cavitation.
Figure 14. scheme of the cavitation presence in the flow at

The latter definition is ofvalid
Figure 13. scheme for rigid
the cavitation cylindrical
presence in the flow at pipe walls.
high opening
low opening
The mixture speed of sound can therefore be calculated as follows:
s
Kmix
cmix = (2.52)
ρmix
where cmix , ρmix the speed of sound and density of the mixture respectively and Kmix
effective bulk modulus of the mixture. The latter is calculated with extensive averaging
of the bulk modulus of gas, liquid and of void fraction (Eq. 2.53). Moreover, the mixture
density is given by Eq. 2.19.
KL
Kmix = K (2.53)
1+α L
KG
−1
Hence, the final formula for speed of sound in two-phase medium is derived:
s
KL
cmix = h i. (2.54)
ρG · α + (1 − α) · ρL 1 + α KKGL − 1
43
The possibility of reaching critical flows even for low flow rates is very high in two-
phase flow mixtures. To demonstrate this, the speed of sound in bubbly air-water flow
function of the void fraction is plotted in Fig. 2.13(b). Both isothermal (polytropic ex-
ponent k=1) and adiabatic (k=1.4) assumptions are demonstrated with relatively small
discrepancies. It is obvious from the shape of both curves that sonic velocity becomes
very small (' 20 m/s) already at 20 % of void fraction. From this diagram, we can con-
clude that sonic velocity of the mixture can be orders of nagnitude lower than of either
of its components. Hence, chocked flow is expected to occur rather often in two-phase
mixtures that develop high velocities in small flow passages such as in the case of safety
relief valve which will be the subject of this study. Another example of 80% drop in
wave velocity in liquid medium due to the presence of only 1% dissolved air is shown in
Fig. 2.13(a).
Wave velocity c, m/s
(a) Propagation velocity of dissolved air in water (b) Speed of sound in bubbly air-water flow at at-
Figure 9.2. The sonic velocity in a bubbly air/water mixture at atmo-
(taken from Wylie and Streeter (1978)) spheric pressure
mospheric pressure for k=1 and k=1.4 (from Bren-
for k = 1.0 and 1.4. Experimental data presented is from
nen (2005)).
Karplus (1958) and Gouse and Brown (1964) for frequencies of 1 kHz (),
0.5 kHz (), and extrapolated to zero frequency().
Figure 2.13: Speed of sound in two-phase flow medium.
be compared with the low frequency analytical results presented here. Note
In this study, interest is stressed
that theon thecorresponds
data presence oftocavitation or flashing
the isothermal phenom- that the heat
theory, indicating
ena and subcritical-critical flows in SRV
transfer whichthe
between will be discussed
bubbles and the in detail
liquid in §2.4.7toand
is sufficient maintain the air
Chapter 6. in the bubbles at roughly constant temperature.
Further discussion of the acoustic characteristics of dusty gases is pre-
sented later in section 11.4 where the effects of relative motion between the
2.4.7 Existing methodology particlesfor
andtwo-phase flow in Also,
the gas are included. SRVthe acoustic characteristics of di-
lute bubbly mixtures are further discussed in section 10.3 where the dynamic
The subject of modeling the discharge
response ofthrough a throttling
the bubbles deviceininthe
are included two-phase
analysis.flow has
been recently studied; models have been proposed by several authors, however until now
no common practice has been widely adopted. The model that has been established as the
most appropriate for most conditions is the Homogeneous Equilibrium
9.3.2 Sonic speeds Modelfrequencies
at higher (HEM), the
so called ω-method. It is a recommended practice from the American Petroleum Institute
Several phenomena can lead to dispersion, that is to say to an acoustic
(API) for sizing relief devicesvelocity
as referred
that isina API RP of
function 520 (2000) or
frequency. in Europe
Among in the
these are other
effects of bubble
national standards; examples are NF anddiscussed
dynamics DIN forinFrance
the nextand Germany
chapter. respectively.
Another For that occurs at
is the change
the United States, the regulation for sizing
higher SRV is
frequencies asdescribed in ASME
the wavelength is noVIII-Div.1 (2010).infinite relative
longer effectively
to the size of the particles. Some experimental data on the effect of the
44 ratio of particle size to wavelength (or κR) was presented in figure 9.1.
Note that the minimum in the acoustic velocity at intermediate volume
fractions disappears at higher frequencies. Atkinson and Kytömaa (1992)
225
Although the international standard concerning safety devices for protection against ex-
cessive pressure in incompressible and compressible fluid is described in ISO 4126-1
(2004), for two-phase flow the latest standard ISO 4126-10 (2010) requires validation and
possibly further implementation. Sizing of SRV according to ISO 4126-10 is described in
a recent reference by Schmidt (2011).
The HEM assumes the same velocity for the two phases that appears to be an accu-
rate hypothesis for a high level of mixing of gas and liquid. In this case, an expansion
coefficient ω is determined and a theoretical critical mass flux is calculated for different
operating conditions. This method can be applied for nozzles, orifices and safety valves;
hence the relief area is settled on. In Design Institute of Emergency Relief Systems, rel-
evant studies have been carried out. In the frame of the DIERS activities, Diener and
Schmidt (2004) and Diener and Schmidt (2005) have investigated the discharge through
nozzles, orifices, SRV and control valves and proposed a boiling delay coefficient to ac-
count for this phenomenon. Thus, the parameter ω is corrected and a new Homogeneous
Non-Equilibrium Diener and Schmidt (2004) model (HNE-DS) is proposed.
At this point, it worths mentioning the activities of the Design Institute for Emergency
Relief Systems (DIERS) on this area of interest. It was a consortium of 29 companies,
under the auspices of the American Institute of Chemical Engineers (AIChE), that devel-
oped methods for the design of emergency relief systems to handle runaway reactions.
DIERS investigates the two-phase vapor-liquid onset / disengagement dynamics and the
hydrodynamics of emergency relief systems.
In the next section (§2.4.7), a detailed presentation of the methodologies of flow dis-
charge in two-phase flow and their limitations presently existing in the open literature is
carried out. Since the models proposed are all valid for non-flashing and/or flashing fluids,
in order to apply the same methodology for the flow conditions studied during this project
(liquid cavitating flow and injected air-water with cavitation flow), a separate paragraph
is dedicated to discussion of modification in HNE-DS method.
Flow discharge through valves and nozzles

The HEM and HNE models are the two main approaches for two-phase flow discharge in
valves and nozzles. Various formulations and versions of these methods are proposed by
several authors. A simplification of the HEM is the ω method that was initially derived by
Leung (1996) and further developed by other researchers. In API RP 520 (2000) a spe-
cial version of the ω method is suggested. Additionally, Homogeneous Non-Equilibrium
model has been proposed by Diener and Schmidt (HNE-DS). Since the formulas used
are highly sensitive on units, it is very important to use the appropriate SI units for their
application. Table 2.2 summarizes the SI units per model. Hereafter the several methods
are demonstrated.
2.4.7.1 Homogeneous Equilibrium Model

In this model, a sufficiently good mixing of the two phases is assumed so that a “pseudo
single-phase” fluid whose properties are a weighted average of each phase can be used. A
mechanical and thermal equilibrium between the two phases is supposed. For a flashing
thermodynamic system, a vapor-liquid equilibrium is involved. For a non-flashing system,
45
Table 2.2: Units for two-phase SRV methodology.

SI units for SRV methods
HEM method
G critical mass flux kg/m2 s
P absolute pressure Pa
ρ density kg/m3
ω method
ρ∗ two-phase mixture density kg/m3
v specific volume m3 /kg
x mass quality -
CP specific heat (liquid) at constant pressure J/kg ◦ K
k ratio of specific heats of the vapor phase (CP /CV ) -
L latent heat of vaporization of the liquid phase at stagnation state J/kg
n pressure ratio (ncrit =P1 /Pcrit , n1 =P1 /P0 ) -
ω expansion parameter -
API RP 520 method
A surface occupied by each phase m2
Ṁ mass flow rate kg/s
Kd discharge coefficient for each phase -
Kb correction factor due to back pressure -
KW correction factor due to back pressure for liquid -
KV correction factor due to viscosity -
M molar mass kg/mol
R gas constant m3 Pa/mol·kg ◦ K
Z compressibility factor -
HNE-DS method
N boiling delay factor -
α parameter HNE-DS -
Subscripts
0 upstream (stagnation)
1 downstream (throating)
L liquid
G gas
LG difference between gaseous and liquid phase properties
crit critical conditions
no mass transfer occurs and this equilibrium implies the same temperature between the
two phases. If, additionally, the hypothesis of no heat transfer between the phases (thermal
insulation) is made, the so-called “frozen flow” is achieved.
Derivation of HEM We consider the nozzle shown in the schematic of Fig. 2.14 which
represents an open thermodynamic system. If we apply the general differential formula-
tions of energy balance for the flow of unit mass of fluid:
u2
!
− dq + dWS + dH + d + gdz = 0 (2.55)
2
46
P0 P1
T0 T1
U0 U1
Z0 Z1
dq dWS
Figure 2.14: Nozzle flow.
where q the heat absorbed by the surrounding, H the enthalpy, WS the work done on
the surrounding, z the height and g the acceleration due to gravity. Assuming isentropic
flow (T dS=dq) and from the first law of thermodynamics (dH=T dS +vdP) we obtain:
dq = dH − vdP (2.56)
where P the absolute pressure and v the specific volume (=1/ρ). If we combine
Eq. 2.55 and Eq. 2.56, the following equation is extracted:
u2
!
dWS + vdP + d + gdz = 0 (2.57)
2
The work done is zero and by replacing d(u2 /2)=udu and neglecting and potential
energy variation (gdz=0):
vdP + udu = 0 (2.58)

The following final relation is obtained:
dP
+ udu = 0 (2.59)
ρ
Rearranging Eq. 2.59 and substituting for the mass flux (G=ρu):
dG dρ
ρ2 = −ρG + G2 (2.60)
dP dP
And when chocked flow is reached, the mass flux yields a maximum (dG/dP=0).
Hence, for frictionless isentropic flow through a nozzle the mass flux G can be derived:
v
u
t ZP1
u
u
u
dP
G = ρ −2 . (2.61)
ρ∗
P0
where ρ∗ the two-phase density along the nozzle path. The integration is made from
the stagnation (inlet) pressure P0 at the nozzle until the throating (outlet) pressure P1 .
The critical flow conditions can then be derived by numerical integration of Eq. 2.61
for chocked flow hypothesis. Estimation of thermodynamic properties of the mixture and
the two-phase flow density in function of the pressure are required. The several versions
of HEM rely upon evaluation of these quantities. This methodology is used to obtain the
mass flux through a nozzle but can also be applied for the case of a SRV.
47
2.4.7.2 The ω method

The ω method, originally developed by Leung (1996), is a special case of the HEM model.
A compressibility factor, called ω is defined (Eq. 2.62); this parameter expresses the
density variation with respect to pressure change from stagnation (denoted with subscript
0) to throat position (denoted with subscript 1).
0→1
∆ ρ P0
ω= · . (2.62)
∆ P ρ0
0→1
Derivation of ω parameter The omega method is an explicit approximate solution of

HEM (Eq. 2.61 is solved analytically). For two-phase flow, v is the specific volume of the
mixture and is given by:
v = xvG + (1 − x) vL (2.63)
If we derive the previous equation, we develop (dvL =0, liquid incompressible) and
divide by dP, we obtain:
dv dvG d ẋ
=x + (vG − vL ) (2.64)
dP dP dP
The specific enthalpy of the two-phase system will be:
h = hL + xhLG = U + Pv (2.65)
This formula for the present case and after differentiating transforms to:
C PL dT + Ld ẋ = dU + vdP + Pdv (2.66)

where L the latent heat of vaporization at stagnation state and CPL the liquid specific
heat at constant pressure. By using the law of conservation of energy at thermodynamic
equilibrium conditions and since vdP=-udu (Eq. 2.58), we acquire:
dx −udu C PL
= − (2.67)
dT LdT L
At critical flow conditions dG/dP=0 which means that G=ρu=ct → udρ+ρdu=0 and
for low values of the void fraction ρ ' ρL → dρ=0. Therefore, du=0 and the following
equation is extracted:
dx CP
=− L (2.68)
dT L
And from use of the Clausius-Clapeyron equation for single component system in
thermodynamic equilibrium:
dT (vG − vL )
= T (2.69)
dP L
With combination of Eq. 2.68 and 2.69 and substitution of term dx/dP in Eq. 2.64 we
obtain Eq. 2.70. Both terms denote flashing flow while in the case of non-flashing flow
48
the term dx/dP (Eq. 2.64) vanishes (x=ct) and only the first term of Eq. 2.70 contributes
to vapor expansion (also called “frozen flow” in some references such as in Diener and
Schmidt (2004)).
Flashing
z }|{
dv dvG (vG − vL )2
= x −C PL T (2.70)
dP dP
|{z} L2
Non-flashing
In order to estimate the specific volume of the gaseous phase vG , the vapor expansion
is assumed to follow an isothermal change of state. Following assumptions are made:
• The variation of the specific volume of liquid with pressure is neglected (vL =vL,0 )
• Ideal gas law is used
• Enthalpy of vaporization is described by a single averaged value (L=L0 )
• These assumptions are NOT valid close to the thermodynamic critical point
According to ideal gas law and by derivation the term vG :
P0 vG,0
P0 vG,0 = PvG ⇒ vG =
P (2.71)
P0 vG,0
⇒ dvG = − 2 dP
P
Equation 2.70 can be rewritten after substitution of dvG and vL =vL,0 as follows:
P v 2
0 G,0
P0 vG,0 P
− vL,0
dv = −x dP − C PL T dP (2.72)
P2 L02
Integration of Eq. 2.72 assuming that vapor quality has a very small variation with
pressure at large qualities (x ' x0 ) and that temperature change is insignificant from
stagnation to throat (T ' T 0 ) yields:
Zv ZP ZP !2
1 CP T0 P0 vG,0
dv = P0 vG,0 x0 − 2 dP − L2 − vL,0 dP (2.73)
P L0 P
v0 P0 P0
Hence, if we also assume small variation of pressure P ' P0 , the following formula is
derived:
Zv ZP !2 ZP
1 vg,0 − vl,0 1
dv = P0 vg,0 x0 − 2 dP + C Pl T 0 P20 − dP (2.74)
P L0 P2
v0 P0 P0
Therefore, we obtain the ω parameter which can be written in the final form:
49
v !2
v0
−1 x0 vG,0 C PL T 0 P0 vG,0 − vL,0
ω= P0
= + (2.75)
−1 v0 v L
P |{z} | 0 {z 0 }
Existing vapor Phase change
The first term of Eq. 2.75 corresponds to expansion of the two-phase mixture due to
existing vapor and the second term to expansion due to phase change (depressurization).
For critical conditions, from Eq. 2.61 and Eq. 2.62, the mass flux is obtained:
r
P0
P1 < ncrit P0 ⇒ G = ncrit (2.76)
v0 ω
where ncrit the critical pressure ratio P1 /Pcrit .
and for the subcritical conditions:
√
−2 [ωln (n1 ) + (ω − 1) (1 − n1 )] P0
r
P1 ≥ ncrit P0 ⇒ G= (2.77)
ω n11 − 1 + 1 v0
where n1 the back pressure ratio P1 /P0 (P1 previously referred as Pb ).

The critical pressure ratio for ω >2 is estimated by Epstein et al. (1983) as follows:
ncrit = 0.55 + 0.217ln (ω) − 0.046 (ln (ω))2 + 0.004 (ln (ω))3 (2.78)
And for ω≤2 the Eq. 2.79 should be solved iteratively:

ncrit + ω2 − 2ω (1 − ncrit )2 + 2ω2 ln (ncrit ) + 2ω2 (1 − ncrit ) = 0 (2.79)
An explicit solution of Eq. 2.79 is proposed by Tyco (2008):
h i(−0.70356+0.014685lnω)
ncrit = 1 + 1.0446 − 0.0093431ω0.5 ω−0.56261 (2.80)
2.4.7.3 HNE-DS (Diener and Schmidt (2004)) method

The authors Diener and Schmidt have proposed a correction factor N of the expansion
parameter ω. This is a boiling delay coefficient to account for the degree of thermody-
namic non-equilibrium at the start of nucleation in low mass fractions of vapor upstream
the fitting.
Derivation of ωDS parameter The rate of evaporation at thermodynamic equilibrium

denoted as dxe /dP in Eq. 2.64 can be transformed to non-equilibrium conditions by the
following formula:
dx dxe
= N (2.81)
dP dP
The parameter N varies from 0 to 1. With the extremes of N=0 at non-flashing (frozen
flow) and N=1 at equilibrium flow. The main parameters to determine the boiling delay
are:
50
• The inlet mass flow quality
• The pressure drop
• The relaxation time between inlet and narrowest cross section
A large non-equilibrium condition is expected in nozzles with small area ratio (and
control valves) since a fast depressurization occurs in short length path (N << 1). The
time for heat transfer between the two phases is very small. In contrast, in safety valves a
more important boiling delay will occur with enough time for heat exchange between the
two phases reaching almost thermodynamic equilibrium conditions.
The boiling delay factor can be written as:
dx
N= (2.82)
dxe
The factor N is the change in mass flow quality between inlet and narrowest cross
section. It denotes the rate of vaporization at a certain depressurization rate. The authors
have used 1300 experimental data points of valves and orifices and a power law fit is
suggested for factor N. Therefore:
N = [xe (Pcrit )]α (2.83)

Integration of this equation can be done by taking the two limits of frozen flow (N=0)
and thermodynamic equilibrium (N=1). Hence, the boiling delay factor covers the whole
range of mass flow qualities (from pure liquid to pure gas). The vapor mass flow quality
at chocked flow is the sum of vapor quality at stagnation conditions and the increase of
vapor content from evaporation due to expansion, called ∆xe (Pcrit ).
xe (Pcrit ) = x0 + ∆xe (Pcrit ) (2.84)

This term can be determined by combining Eq. 2.68 and Eq. 2.69, using Eq. 2.71 for
dvg and then integrating from P0 to Pcrit :
C PL (vG − vL )
dxe = − · T dP
L0 L0
PRcrit
Eq. 2.71+ ZPcrit P v
0 G,0
− v

P0 P L
→ ∆xe (Pcrit ) = −C PL T dP (2.85)
L02
P0
!
C PL T 0 P0 vG,0 − vL,0 P0
∆xe (Pcrit ) = ln
L02 Pcrit
Therefore, the final form of boiling delay factor is given by:

!!α
C PL T 0 P0 vG,0 − vL,0 P0
N = x0 + ln (2.86)
L02 Pcrit
Therefore, the parameter as defined from this model becomes:
51
!2
x0 vG,0 C PL T 0 P0 vLG,0
ωDS = + ·N
v0 v0 L0
where the boiling delay factor N is given by:
" ! !#α
vLG,0 1
N = x0 + C PL T 0 P0 ln (2.87)
L02 ncrit
The exponent α is approximated by following equation of critical mass flux (from
Henry and Fauske (1971)):
q
ωln ncrit
1
− (ω − 1) (1 − ncrit ) r 2P
0
Mcrit = h i . (2.88)
ω ncrit − 1 + 1
1 v0
Hence, the recommended values from the authors (also in accordance with Henry and
Fauske (1971)) for α are:
α = 3/5 For orifices, control valves and short nozzles,

α = 2/5 For safety valves and control valves (high lift),
(2.89)
α'0 For long nozzles and orifices with large area ratios.
The authors have validated the accuracy of the model against experimental data from
Lenzing and Friedel (1996) and Lenzing and Friedel (1998). A very good description of
the HNE-DS method for initially subcooled TP flow is given in the quite recent reference
by Schmidt (2007).
2.4.7.4 Modified CF-HNE-DS (Diener and Schmidt (2004)) method

Due to the lack of existing methodologies which can predict the discharged mass flux
through injected air-water with cavitating flow (CF), in current study, the HNE-DS model
with modifications has been applied. Hence, considering air-water without presence of
cavitation, only the first term of ωDS is calculated using the measurements of air an wa-
ter flow. When influence of cavitation is taken into account, the second term of ωDS is
computed using the properties of the two fluids (air and water) and estimation of the ho-
mogeneous mixture density with Eq. 2.18. The boiling delay factor can then be calculated
with Eq. 2.87 provided that exponent α is known. As a first approximation, the values pro-
posed from the authors (3/5 and 2/5 for short nozzles and safety valves respectively) are
checked. As it will be presented in Chapter 6, a very good agreement of the model is
found with fitting parameters 3/5 and 2/5 for industrial and transparent SRV accordingly
revealing that the former behaves as a short nozzle while the latter as safety valve.
Thus, based on the two tested cases, we conclude that air-water flow with
cavitation simulated the behavior of flashing liquid and for these specific
conditions it can be modeled with modified method of Diener and Schmidt
(CF-HNE-DS) although further validation of this observation is required.
52
2.4.7.5 API RP 520 (2000) method

The recommended practice (RP) from American Petroleum Institute (API RP 520 (2000))
applies to the sizing and selection of PRVs used in refineries and related industries for
equipment that has a maximum allowable working pressure of 15 psig [103 kPag] or
greater. The pressure relief devices covered in this recommended practice are intended to
protect unfired pressure vessels and related equipment against overpressure from operat-
ing and fire contingencies.
For the overall mass flux, using the drift-flux separated two-phase flow model of Wal-
lis (1969) (developed in Appendix I.2), one can obtain (Tran and Reynolds (2007)):
" #−1
xAG (1 − x) AL
G= + (2.90)
MG ML
where the total flow area is the sum of the area occupied by the liquid and the one
occupied by the gas: A=AL +AG . The mass fluxes of each phase GG (MG /AG ) and GL
(ML /AL ) are given respectively by:
r s !(k+1)/(k−1) 
M 2
GG = Kd Kb P0
 
k (2.91)
k+1

RT Z
and
G L = Kd KW KV 2ρL (P0 − P1 )
p
(2.92)
where Kb the correction factor due to back pressure for gas, M the molar mass in
kgmol−1 , R the gas constant in m3 Pa ◦ K−1 kg−1 mol−1 , Z the compressibility factor (Z=1
for ideal gas), KW the correction factor due to back pressure for liquid and KV the correc-
tion factor due to viscosity. Two additional important parameters in SRV can be defined
according to API RP 520 (2000):
• Effective discharge coefficient is a nominal value used with the effective discharge
area to calculate the minimum relieving capacity of a PRV estimated with the pre-
liminary sizing equations given in API RP 520 (2000).
• Rated discharge coefficient is the discharge coefficient determined in agreement
with the required for the specific application standard or code and is used with the
actual discharge area to calculate the rated flow capacity of a PRV.
2.4.7.6 Comparison of models

A comparison of different two-phase flow models for various hypothetical relief scenarios
is presented in Fig. 2.15. In this plot, increasing number of relief vessel corresponds to
decrease of the the set pressure which was the main varying parameter. The HEM tends to
estimate the highest relief areas while the API RP 520 (2000) predicts considerably lower
relief areas, hence HEM is a more conservative methodology compared to API RP 520
(2000). Separated flow empirical methods result in intermediate values of discharged
area. We should point out that this study is not a validation since no experimental values
are provided. Finally, a summarizing matrix of the existing methods for SRV sizing in
two-phase flow and their application range and assumptions is presented in Table 2.3.
53
2 Theoretical background-Literature review
Limitations-
Limitations-
Model References Formulas
assumptions
P1
dP
−2 ∫
•Homogeneous mixture
HEM - G=ρ
P0
ρ * •Isentropic flow (reversible
adiabatic)
•Mechanical & thermal
⎧ −2 ⎣⎡ω ln ( n1 ) + (ω − 1)(1 − n1 ) ⎦⎤ ⎫ equilibrium
⎪ Subcritical ⇒ P1 ≥ ncrit P0 P0 ⎪
⇒ G= Non-flashing
P
⎪ ⎛1 ⎞ v0 ⎪ 2
•Ideal gas behavior
⎪ ω⎜ − 1⎟ + 1 ⎪ x0 vG ,0 CPL T0 P0 ⎛ vLG ,0 ⎞ •Heat of vaporization & heat
ω method Leung [1996] ⎨ ⎝ n1 ⎠ ⎬⇒ ω = + ⎜ ⎟ capacity of fluid are constant
⎪ ⎪ v0 k v0 ⎝ L0 ⎠

throughout the nozzle
⎪ Critical ⇒ P1 < ncrit P0 ⇒ G = ncrit
P0 ⎪ Flashing
⎪ v0ω ⎪ •Vapor pressure &
⎩ ⎭ temperature follow the
Clapeyron equation
2
x0 vG ,0 CPL T0 P0 ⎛ vLG ,0 ⎞
ωDS = + ⎜ ⎟ ⋅N α =3 5 Orifices, control valves and short nozzles
v0 k v0 ⎝ L0 ⎠
α •Equilibrium and non-
⎡ ⎛v ⎞ ⎛ 1 ⎞⎤
HNE D-S Diener and Schmidt [2004] N = ⎢ x0 + CPl T0 P0 ⎜ LG2,0 ⎟ ln ⎜ ⎟⎥ α =2 5 Safety valves and control valves (high lift) equilibrium conditions
⎣ ⎝ L0 ⎠ ⎝ ncrit ⎠ ⎦
N = 1 ⇒ Equilibrium
α ≈0 Long nozzles and orifices with large are ratios
N ≤ 1 ⇒ Non-equilibrium
( k +1) ( k −1) ⎤
1 M ⎡ ⎛ 2 ⎞
G= , GG = K d K b P0 ⎢k ⎜ ⎟ ⎥
⎡ xA (1 − x ) AL ⎤ RTZ ⎣⎢ ⎝ k + 1 ⎠ ⎦⎥
•Based on Wallis [1969]
API API RP 520 [2000] ⎢
G
+ ⎥ model
⎣ MG ML ⎦
GL = K d KW KV 2 ρ L ( P0 − P1 )
Table 2.3: Summarizing matrix of two-phase flow methodology in SRV.
54
54
2.4.8 Literature review on SRV

The number of bibliographic references concerning the study of Safety Relief Valves
(SRV) in the last decades has considerably increased. Most of the articles concern ex-
perimental investigations and theoretical modeling of single-phase compressible or in-
compressible flow while fewer studies deal with two-phase flow. Lately, numerous are
the research projects focused on CFD simulations applied in single-phase SRV while for
two-phase flow a remarkable absence of references can be stated. A clear advantage
of numerical simulations is that they can validate of models with significantly reduced
cost compared to relatively expensive valve tests. Since main interest of this study is fo-
cused on experimental research, only small part of existing CFD bibliography is presented
herein.
Figure 2.15: Comparison of discharged relief area calculated with different two-phase
flow models (from Tran and Reynolds (2007)).
2.4.8.1 Experimental studies-Modeling

In open literature, there is a lack of experimental data concerning the sizing of SRV. This
has led many researchers to attempt a theoretical modeling of the phenomena taking place
when a SRV opens. Both flashing and non-flashing conditions should be considered. Boc-
cardi et al. (2005) have studied steam/water flashing flow. Varying parameters are vapor
quality, inlet pressure, mass flow rate and back pressure. Whilst giving an overview of the
available models, they state that the ω method (HEM) underestimates the maximum flow
rate discharged through a safety valve. Hence, they recommend a two-phase discharge
coefficient Kd correlation function of the back pressure ratio n1 and inlet pressure P set .
The proposed discharge coefficient is a linear correlation:
Kd = mn1 + q
55
Parameters and application range of this correlation are reported in Table 2.4. In this
matrix, Kd (1) represents the discharge coefficient for pressure ratio equal to unity and m,
q, r2 are the regression parameters.
Table 2.4: Discharge coefficient Kd proposed by Boccardi et al. (2005).

P set [MPa] m [-] q Kd (1) r2
0.5 -0.882 1.610 0.728 0.365
0.75 -1.214 1.850 0.636 0.802
1 -1.601 2.186 0.586 0.864
1.25 -1.284 1.862 0.578 0.857
1.5 -1.155 1.728 0.574 0.891
1.75 -1.055 1.616 0.561 0.830
Lenzing et al. (1998) carried out measurements in commercial full lift SRV for flash-
ing and non-flashing conditions. A comparison between the measured discharged mass
flux and the calculated with HEM, ω method, Homogeneous Frozen Flow (maximum dis-
equilibrium) by Nastoll (1985) and the practice by Goßlau and Weyl (1989) proved that
the agreement is satisfactory in a limited application range. A weighted homogeneous
two-phase discharge coefficient (Eq. 2.93) is proposed for single-component flashing flow
to extend the Homogeneous disequilibrium model of Henry and Fauske (1970) and the
predictive accuracy is considerably improved.
KdT P = αKdG + (1 − α) KdL (2.93)

where KdL , KdG and KT P are the discharge coefficients of liquid, gas and two-phase
mixture respectively. Lenzing et al. (1998) compared experimental results for steam-
water mixture with predictions from the formula using HEM and they concluded that
considerable underestimation of the flow rate is noticed for all vapor qualities.
In their paper, Darby et al. (2001), present a review of the most commonly used models
for two-phase flows in valves, nozzles and tubes. Moreover, they discuss the importance
of thermodynamic conditions (frozen, equilibrium and non-equilibrium flow) and length
of nozzle on the choice of the appropriate model. A large experimental flashing and non-
flashing database is used for this purpose.
Shannak (2009) has adopted a methodology to estimate the loss coefficient of gas-
liquid mixtures in SRV. By using three different types of valves to obtain experimental
data and combining these data with Weisbach (1845) equation, the author considered flow
contraction correlation, viscosity correction factor, mixture Re number and safety valve
geometry to extract the associated formula for pressure loss coefficient (2.94). Compari-
son between a large number of sampling data in literature and prediction from the model
gives satisfactory results with a maximum standard discrepancy from the data of 27 %.
!
h dinlet
K(m,S V) = f Cc(m,S V) , Re(m,S V) , η, , , S V(T ype) (2.94)
d0 doutlet
Analytically, the correlations used in the model proposed from Shannak (2009) are
given below2 :
2
Subscripts m, SV, 0 indicate the mixture flow, Safety Valve and narrowest flow path respectively
56
!−1 
 Cc(m,S V) 0.01
 ! !0.05 !!0.2
1 d inlet h
= A  + 0.001

K(m,S V) 1− (2.95)
η

Re(m,S V) doutlet d0 
with the different terms of Eq. 2.95 given by:
Cc(m,S V) = 0.52 (1 − x)0.05 + 0.51 (x)0.05 , (2.96)

x2 + (1 − x)2 (ρG /ρL )
Re(m,S V) = , (2.97)
x2 /ReG + (1 − x)2 /ReL (ρG /ρL )
 !2 −0.038 " #−0.087
1.24 d0  Lnozzle
η= √

2 −  (2.98)
170/ReL + 0.89 dinlet d0
where A depends on the type of valve and is equal to 0.4 (SRV type Leser) and 1.2
(SRV type Bopp & Reuther and Sempell), Cc the contraction coefficient, Lnozzle the length
of the inlet nozzle in m, dinlet and doutlet the inlet and outlet diameters of the PRV respec-
tively in mm and h the valve lift in mm. Bopp & Reuther
Föllmer and Schnettler (2003) studied the force applied on the
The test valves for steam were checked on this test facility in the same way valve diskwith
(called flow
air. The
forcesubsequent
from nowverification
on) for different
during set
the pressures and lift with
tests in America values. “Flowforce-characteristic-
saturated steam confirmed our
measurement”
measuring is used to
results withcharacterize
regard to theflowdifferent parts ofi.thee.PRV
and function, steamandispossible optimiza-
also covered as
tion compressible
of these parts medium.
can be concluded (method originally developed by Bopp & Reuther
This large-scale
manufacturer test facility
in Schnettler has beenFinite
(1994)). described with all
Element its possibilities
Method (FEM) in [5].
calculations together
with stability computations are performed on the shape of the valve body. An example of
Before looking for a suitable spring in order to solve the task opening and closing within
flowforce characteristic
given limits is task
there is the plotted in Fig. 2.16.
of developing Force
a flow is increasing
geometry which is linearly with pressure
able to furnish optimal
and asymptotically with valve lift. This is due to pop-opening action of the valve for the
flowforce gradients.
smallest lifts as it was explained in §2.4.2.
Force [N]
Pressure [bar]
Lift [mm]
Figure 2.16: Flowforce characteristic of a PRV taken from Föllmer and Schnettler (2003).
Figure 3: Flowforce characteristic of a safety valve
57
For this purpose Bopp & Reuther developed the method of "flowforce-characteristic-
measurement" [6]. This method permits the evaluation of the quality of the flowforce
performance gradients over a large pressure range (figure 3). Local discontinuities which
usually remain undiscovered in case of individual function tests, if a test is not by accident
carried out within the hazardous pressure range, are immediately discovered with this
method.
Together with the "flowforce-characteristic-measurement" the flow behaviour of the safety
Dossena et al. (2002) have studied the effect of back pressure valve size on the flow
capacity of SRV. Back pressure will result in discharge coefficient and thus flow rate re-
duction. The reason for this phenomenon is possibly inadequate disk lift or appearance of
subsonic flow along the flow path. The authors state that the effect of the aforementioned
parameters on different valves can be different due to the different geometrical scaling of
the valves according to API 526 (2002). After their tests, they have noticed a big differ-
ence on the flow discharge capacity when applying back pressure. Therefore, every valve
should be tested under different expansion rates to extract its performance. Finally, they
proposed an experimental correlation to predict the non-similarity effects.
Dossena et al. (2004) investigated the high pressure fluctuations on the valve disk in
the opening region for three different types of PRVs. A resulting vibration in the upstream
piping (even in the protected system above a certain length) is observed. The analysis of
vibration occurrence shows that the phenomenon is driven by a harmonic self-excited
motion of the valve disk coupled to a standing wave system in the connecting pipe. No
damage on the nozzle seat is observed due to vibrations contrary to what happened on the
stem-disk couple.
Chabane et al. (2009) carried out an investigation in conventional SRV with built-up
back pressure. For back pressure higher than 30 % of the set pressure, balanced bellows
configuration is recommended to avoid vibration and chattering of the valve disk. How-
ever, the authors stress the attention on the fact that even though, according to common
practices, for Pb lower than 10 %, conventional SRV can be used, fluttering which can lead
to damage of valve components is possible. In the article, a 1D model is developed with
inlet and outlet conditions of SRV given by thermodynamic 1D and wave propagation
model. Flowforce conditions are provided by CFD calculations.
2.4.8.2 Numerical studies in single-phase flow

Song et al. (2010) analyzed the dynamic behavior of a pressure relief valve with three-
dimensional Computational Fluid Dynamics (CFD) calculations. The system reservoir-
PRV is modeled and compressible flow from valve opening until re-closure is investi-
gated; moving meshes are used for this purpose. An example of the constructed mesh is
shown in Fig. 2.17(a) and in Fig. 2.17(b) a plot of the velocity streamlines in the valve
is depicted. In the lateral diagram, higher velocities are remarked close to the valve disk
which represents the smallest flow passage section. Furthermore, direction of the stream-
lines shows a formation of “hat” shape flow caused by the conical form of the valve disk.
The authors have observed a severe oscillation during re-closure process of the valve
which can result in physical damage of the moving parts. Thus, a modification of the
system disk-disk holder and spring of this valve is proposed in order to prevent vibration.
Vallet et al. (2010) used Reynolds-Averaged Navier-Stokes (RANS) k-ω Shear Stress
Transport (SST) modeling to simulate single-phase water flow in 1 1/2 " SRV with a 3D
hybrid unstructured mesh. Hydrodynamic force and flow coefficient CV for different valve
openings is investigated. Stationary, axisymmetric flow for the valve is noticed. Although
cavitation is not reproduced in the model, comparison of computations with experimental
results is satisfactory. Therefore, it is concluded that single-phase CFD simulations can
provide important information in cavitating SRV. Figure 2.18 shows velocity iso-surfaces
for different inlet velocities and high P set up to 3 MPa. Axisymmetric flow upstream and
58
(Nd.s/Em.s) of each domain. A pure structural grid (hex mesh)
the region between the disc and the adjusting ring, the flow bore area rather than
was generated for the four sub-domains. Minor simplifications,
accelerated further to a maximum value of approximately Mach During this period, th
such
2.4as Industrial
neglecting thevalves
fillets and chamfers, were made to formation of a recirculating zone caused a reduction in
2. The 59 As the air exhausted
ensure all the domains could be meshed with a pure structural
the flow area, and thus a rise Valve
in the part
velocity of the fluid with a inside the vessel d
grid. consequential drop in pressure. Consequently, the lift
As the flow moved out from the gap, the main flow force decreased gradu
diverged and was deflected outward toward the exit. In the ValveAt 1.375 s, sinc
Nd.s/Em.s:
opposite direction, the fluid flowed upward due to the valve disc started t
Outl
13598/11980
Sub-domian 3: obstruction of the valve body. As a result, a recirculating flow opening. However,
Vessel:
et
Central part of Valve formed near the inside of the valve body. The corresponding suddenly downward
: ope
10.7bar
pressure distribution also revealed the main pressure drop in the force acting on the va
Nd.s/Em.s:
seat region. As the valve re-closed, the small amount of lift was a resultant forc
n
29784/25480
Sub-domian 4: separated the whole field into two pressure fields. change was very slig
Valve vent Significant changes in the flow direction resulted in the 8(c), both the veloc
Pipe
formation of recirculation zones and modified the minimum reversed, but the av
flow areas, thereby affecting the pressure distribution Enlarged and the Each reversal in the
force on the disc. Careful consideration of these valve partand the
regions on the disc more unb
Sub-domian 2: flow pattern could help produce an improved valve design. forces increased the
Connected pipe Fig. 4. Final CFD model and boundary conditions.
the magnitude of the
Nd.s/Em.s:
the lift decreased.
9576/8250 d) Variable time steps The intensity o
A PSV commonly undergoes three stages from the opening
increase continually
state to the re-close state: rapid open (pop), stay at air the and
highest
the resistan
lift, and re-close. As shown by many previous simulations, maximum magnitud
these three stages require very different durations:decreased. the rapidSince the
open can usually be finished in 0.05 s; the second stage time takes
of thea PSV w
few seconds to minutes based on the overpressure invelocity of the disc a
the vessel;
Nd.s/Em.s: the valve closed rapi
26104/23850 and the re-closure stage usually takes several minutes. Hence,
to save the computational time and account for flow gradients had no significant in
Sub-domian 1: during this short t
and disc motion to ensure convergence, a variable time step
Pressure vessel compression of the
(scale: 0.1)
was used at the different stages rather than a fixed time When step.the overpres
Four time steps were used throughout the process. A time played
overpressure
(m)
step of 1e-6s was chosen in the first 200 steps (2e-4 s). The
contrast, when the
Fig. 3. Structural grids of the four sub-domains. purpose Fig.of6. this settinginwas
Streamline to ensure calculation convergence.
the valve. compression of the s
(a) Mesh of coupled PRV-vessel model (b) Velocity
After streamlines
successfully in PRV forthecompress-
obtaining entire flow field,thea slightly
remaining pres
b) ible flow
Motion larger
of the time
disc andstep of
forces 2.5e-4
on the s was
disc used, which was Therefore, to
sufficient no oscillat
c) Boundary conditions
A small
the gap/lift
After generating the four sub-domains, the whole domainFigures 8simulate and 9 show the diagrams
openingofprocess.
the dynamicsThen, (including
from 5e-2 s (when
mm) was left to ensu
displacementvalve
and velocity)
disc hadof(2010)
the valve its
achieved andlargest
the forces
lift),onaitcomparably
as a large time
was made by connecting them Figure 2.17:viaResults
together obtained
three general
functionby
grid Song
of time, et al.
respectively. Figures 8(b) and (c) are enlarged s on, the valve disc
interfaces (GGIs) as shown in green region in Fig. 4 [20]. In of Fig.step
views
of 1e-3s was used since the valve disc continued to be in
resultant force was
particular, there is no inlet in this case; instead, the flow source its8(a) showing
largest a 20
lift for ms opening
a period of time,stageandandthe 1.0fluid
s flowing during
oscillation of the for
period of the re-closure process, respectively. Compared with
(set pressure and overpressure of 7%) was provided byFig. this period was relatively easy
the7, we note that once the valve had opened, due to the to simulate. From about
observed.s This
1.3 to was p
aroundvessel.
additional the disk is observed
Compared and thus,
with a given for simulating
inlet pressure, this flowthis
sudden the full re-closure
valve,
increase in theaseat2Dstage,
mesh
region, a the
smaller
can be
pressuretime step of 4e-4closer
used.
build-up s wastoused the bottom
more closely represents the actual situation, which allows the
®overcame
again
the to accurately
spring force which capture
causedthethedisc discmovement
to lift andfrom fluidit when
field the disc
Moncalvo et al. (2009) used CFX-Flo substantially
pressure inside to decrease according to the flow condition
commercial during
at pop.CFD
the software
closure
Within process.
about 6 ms,toInstudy valveair
theaddition, aflow
parallel computation
rapidly
through
through two small
the valve. The setvalves
pressure(Leser
was 10 type
bar, so441 DN25/40was
the given andperformed
type 459to DN15/25).
save computational
They time.
dis- The task was
pressure in the vessel sub-domain was 10.7 bar. The outlet divided into four pieces, and each piece communicated via the
cuss the impact of the flow volume discretization (refinement of the
Message Passing grid)(MPI)
Interface and library.
turbulence
condition was set to be open, which allowed flow in both 5
modeling in their computations. It is found that
directions. For the wall specification, most of the surfaces in ω-based SST and standard k-ω turbulence
themodels givesetthe
model were best
to be reproductive
stationary except theaccuracy
3-D regionswith 5. RESULTSmeasurements
of experimental AND DISCUSSION in terms of
air mass flow rate; the more complex Reynolds stress model brings no substantial im-
provement in estimation of the mass flow rate. Finally, 4
ideal gas behavior is tested against
Copyright © 2010 by ASME
the more complex hypothesis of air being an ideal gas mixture of oxygen and nitrogen (be-
having like Redlich-Kwong gases). The latter assumption is proved to give less precise
results than ideal gas behavior.
The PhD thesis of Beune (2009) is concentrated on high pressure safety valves (wa-
ter and nitrogen up to 45.3 MPa). For such high pressures (or low temperatures), real
gas effects are significant hence affecting sizing methodology and opening characteristics
of the valve. Applicability range and validity of different real-gas methodologies (ideal,
real-average and real-integral) is discussed. Safety valve field tests are carried out in
order to provide information on physical effects on safety valve and to deliver an exper-
imental database used in CFD computations with commercial code ANSYS CFX. The
author concludes that “...only dynamic simulations can realistically model the opening
characteristic, because these force peaks have not been observed in the static approach.
Furthermore, the valve geometry can be optimized without sharp edges or cavities so that
redirection of the flow will result in gradual flow force changes...”.
59
Pset~up to 0.3 MPa
Figure 2.18: Velocity iso-surfaces for water flow in PRV taken from Vallet et al. (2010).
2.5 Conclusions for Chapter 2

In this chapter, the common concepts and terminology used in the next chapters are re-
ported. The main single- and two-phase flow parameters used for singularities and safety
relief valve studies are introduced. Several flow patterns in horizontal and vertical straight
pipe flow are discriminated since major influence of flow topology on pressure drop is ex-
pected; inlet and outlet flow conditions are crucial for study of singularity (horizontal
flow) and SRV (horizontal/vertical flow). The most common flow pattern maps for hori-
zontal and vertical (upward and downward cocurrent) flow are presented and the validity
of the latter is confirmed with flow visualization. Vapor-liquid flow (a commonly met
regime in SRV) is analyzed and flow patterns are distinguished as well.
Next part of the chapter is dedicated to modeling of pressure head loss constituting
of three terms; pressure drop due to friction, gravity and acceleration. Therefore, dif-
ferent approaches used to reproduce two-phase flows in straight pipes are recalled; the
homogeneous and the separated flow models. The first one presupposes a very good mix-
ing since equal velocities and temperatures between the two phases is assumed and it is
the most commonly used concept in two-phase flow. In the second approach, the sepa-
rated flow Lockhart and Martinelli (1949) and Martinelli and Nelson (1948) models are
considered for air-water and steam-water mixtures respectively. More complex situation
is faced when two-phase flow occurs in singularities. The two relatively simple homo-
geneous flow models to predict the pressure drop in sudden expansion and contraction
geometries used during this study are Jannsen and Kervinen (1966) and Chisholm (1969)
and are illustrated in next sections of the chapter.
Furthermore, the various types of safety valves that can be found in the industry are
classified and the vocabulary used when these valves are operating is reported. Details on
the working principle of SRV are given and the role of adjusting ring, valve disk and noz-
60
2.5 Conclusions for Chapter 2 61
zle geometry are explained. The definition of discharge coefficient (ratio of experimental
to theoretical mass flow rate through nozzle) for compressible and incompressible fluid
is given both in critical and subcritical regime. Critical flow often occurs in SRV under
compressible flow when sonic velocity is locally reached in the smallest flow passage and
the flow rate is not increasing with further decrease of the pressure downstream. This
phenomenon can often happen in two-phase flow since the speed of sound in two-phase
medium can be orders of magnitude lower than in each constitute. Cavitation is also very
frequent phenomenon in safety valves and it happens when the pressure drops locally
sufficiently low that vapor bubbles are formed.
At the last segment of this chapter, analytical derivation and limitations-assumptions
of the most important methodologies used to predict the mass flux through nozzles in two-
phase flows, i.e. the Homogeneous Equilibrium model (HEM), the ω method, the HNE-
DS method and the API520 model, are provided. For HEM, homogeneous isentropic and
reversible adiabatic flow is assumed. The ω method is an explicit solution of HEM making
the additional hypotheses of mechanical and thermal equilibrium, ideal gas behavior, that
the latent heat of vaporization of the liquid is constant throughout the nozzle and that
vapor pressure and temperature follow the Clapeyron equation. HNE-DS method is an
extension of ω model which is valid both for equilibrium and non-equilibrium conditions,
taking into account the boiling delay between the two phases. The proposed API520
method is based on Wallis (1969) drift-flux separated flow model.
Finally, a brief bibliographic review on research in singularities (contraction and ex-
pansion) and safety relief valves (experimental and numerical) is presented at the end of
each of the two sections.
61
Part II
Measurement campaign
3 Experimental techniques-facilities
Four main experimental facilities have been built and operated during this study. In Table
3.1, the different types of study and dimensions of tested components are reviewed for
each configuration.
Table 3.1: Presentation of the four main experimental facilities.

Facility Test Dimensions Angle [ ◦ ] Type of test
LUCY II Contraction, DN32/40 9 Pressure-
expansion visualization-
optical probe
AGATHE Contraction, DN40/65, 5, 8, 15 Pressure-
expansion DN65/80 visualization
LUCY III Safety Valve 1 1/2" G 3" - Visualization-
optical probe
AGATHE II Safety Valve 1 1/2" G 3" - Pressure-
visualization
The project is split in two phases; the first one concerned the study of geometrical
accidents with two installations AGATHE (large scale) and LUCY II (small scale). In the
second part of the study, a SRV of type 1 1/2" G 3" is investigated. The initial AGATHE
setup is modified to perform pressure measurements in SRV transparent model (AGATHE
II). Additionally, tests are carried out in the industrial valve in air, water and two-phase
flow. As a final step, optical probe measurements are performed in SRV facility LUCY
III. A detailed description of each setup is given in the next paragraphs.
A synthetic table of all the tests realized is illustrated in Fig. 3.2.
3.1 Characterization of facilities for singularities

Experimental horizontal closed loop installations are built at different scales to investigate
the pressure evolution along the upstream part, along and downstream the singularity. The
purpose is to study the influence of the geometrical accident on the pressure drop and ex-
tract correlations to predict the singular pressure change for different flow conditions. The
facilities are initially operated in single-phase water flow and then air is injected before
the singularity and measurements are acquired for both cases. The results are compared
in order to quantitatively determine the influence of the second phase. Furthermore, for
In table ** a synthetic table of the tests realized is presented. The different
66 geometrical characteristics, the type of experimental investigation,
3 Experimental numerical
techniques-facilities
simulations and modelling (via a correlation that is deduced) are pointed out.
SYNTHETIC TABLE OF TESTS

Experimental results
Geometry σ [-] Scale Dimensions Angle Correlation
Pressure Visualization Probe
Smooth
9 9 8 1.56 Small DN40/32 9° 9
contraction
Smooth
9 9 9 0.64 Small DN32/40 9° 9
expansion
Smooth
9 9 8 0.43 Large DN40/65 5° 9
expansion
Smooth
9 9 8 0.43 Large DN40/65 8° 9
expansion
Smooth
9 9 8 0.43 Large DN40/65 15° 9
expansion
Sudden
9 9 8 expansion
0.43 Large DN40/65 90° 9
Smooth
9 9 8 0.65 Large DN65/80 8° 9
expansion
Sudden
9 9 8 0.65 Large DN65/80 90° 9
expansion
Smooth
9 9 8 2.34 Large DN65/40 15° 9
contraction
Sudden
9 9 8 2.34 Large DN65/40 90° 9
contraction
Safety
9 9 9 0.25 Large 1 ½ ʺ G 3 ʺ - 9
valve
Table 3.2: Summarizing matrix of experimental tests.
the flow structure examination, optical probe is used in selected configuration for two-
phase flow to measure the local void fraction, bubble velocity and diameter. Qualitative
results concerning the flow regime, possible stratification due to buoyancy and other phe-
nomena are observed through flow visualization performed with a high-speed imaging in
transparent test sections.
In Figure 3.1, the two different types of expansion geometries tested are presented.
Figure 3.1 A) shows the divergent pipe with opening angle α and Fig. 3.1 B) the sudden
OLD The normalized reattachment length L/d, noticed in Fig. 3.1, denotes the
expansion.
eventual recirculation zone. In the case of convergence geometry the flow direction is
inversed and a contraction region can be observed; a vena contracta is formed in the pipe
downstream the singularity. 1
3.1.1 Small scale

In this part, the study is concentrated in progressive contraction and expansion of 9 ◦ open-
ing angle, surface area ratio σ=0.64, internal pipe diameters of 32 and 40 mm, velocities
up to 6 m/s for water, 2 m/s for air and volumetric qualities of air until 40 %.
A general schematic of the facility is presented in Fig. 3.2. A centrifugal pump (1)
with a maximum capacity of 18 m3 /h is sucking water from a reservoir and the rotational
speed of the motor of the pump is controlled with a frequency inverter. During the exper-
iments, an air release valve (9) connected to the tank is open to the atmosphere to remove
air from the setup. A by pass valve (8) is used to prevent facility from water hammer
phenomenon. An electronic flow meter is employed to measure the water flow rate (3);
the maximum measuring ability of the flow meter is 18 m3 /h. The highest desired flow
66
3.1 Characterization of facilities for singularities 67
A)
Flow α D
d
Reattachment length-L/d
B)
Flow
D
Figure 3.1: A) Progressive ex-
d
pansion of different opening
angles-reattachment length L/d.
B) Sudden expansion-
Reattachment length-L/d reattachment length L/d.
rate is around 11 m3 /h. The setup consisted of a calming length (5) of 60 upstream pipe
diameters to assure a fully developed flow after the bend and before the singularity. Close
to the test section, the injection of air is performed
Figures for thesis VKI
through a gas injector (4). A regulation
valve for
von Karman Institute (2)Fluid
controls the air that is
Dynamics-Centre supplied
Technique desfrom a compressor.
Industries MécaniquesThe air flow rate is measured
Exact
by position offlow
an electronic injection??
meter (6) for up to 20 l/min and for highest flow rates (until 11.5 l/s)
Pressure
with a rotameter (7). A draining valve is also
Small located at the bottom of the reservoir.
scale
9
Inverter
2
Water tank 1 Pump
2 Regulation valve
1
8 3 Electronic water flow meter
3
Water discharge 4 Air injector
5 Calming length
6 Electronic air flow meter (20 l/min)
4
7 Rotameter (1.5 l/s)
5 8 By pass valve
Test section 9 Air release
2 2
7 6
Compressed air
Figure 3.2: Experimental small scale LUCY II facility.
In order to produce uniform bubbly flow, two different gas injectors were manufac-
tured as shown in Fig. 3.3. The injector producing smaller bubbles having four rows of
metallic tubes with several (116) tiny orifices of 0.5 mm diameter (Fig. 3.3(a)). A detailed
picture of the second gas injector is shown in Fig. 3.3(b) with four rows as well and a
total of 28 holes of 1 mm diameter each.
After investigating the influence of the gas injector used, presented in §4.3, no sig-
nificant difference in the structure of two-phase flow obtained is found. Therefore, the
67
68 3 Experimental techniques-facilities
injector of 1 mm orifices is used for all the tests.
(a) Gas injector of 0.5 mm holes (b) Gas injector of 1 mm holes
Figure 3.3: Gas injectors used in small scale setup.
3.1.1.1 Pressure measurements

A picture of the test section for pressure measurements is shown in Fig. 3.4(a). Pressure
taps are placed every 1d upstream the singularity. The exact position of the pressure taps
is indicated in Fig 3.4(b). Since pressure is measured only in convergent section for this
case, closest pressure taps upstream and downstream the singularity are placed at one
diameter distance. This appears adequate, considering that established flow before and
after the convergence is expected at ≥ 1d distance.
3.1.1.2 Optical probe-visualization

Modification of the setup to perform optical probe experiments is demonstrated in Fig.
3.5. This configuration allows measurements of void fraction, bubble diameter and veloc-
ity upstream and downstream a divergence section. The surface area ratio is σ=0.64 for
32 and 40 mm inner diameter of the pipes upstream and downstream respectively. The
opening angle of the enlargement is 9 ◦ . The two-phase mixture consists of air and water.
Special type of flange is adapted to the pipe to acquire the void fraction, bubble size
and velocity with a dual optical probe in a location situated at 6 tube diameters upstream
and downstream the singularity. Profiles are obtained in the horizontal and vertical plane
with directions as indicated in the schematic. Additional profiles are taken at several lo-
cations after the air injection with the purpose of identifying any stratification of the flow.
Detailed drawings of the test section are given in Fig. B.1. The test section consists of a
box made of Polymethyl Methachrylate (PMMA) which is filled with water to decrease
the optical aberration effect as demonstrated in Fig. 3.5.
3.1.1.3 Test conditions small scale

The two fluids tested are air and water in isothermal conditions. The volumetric quality
of air varies from 0-40 % and uniform dispersed bubbly flow is the dominant regime. The
68
Small scale
LUCY II drawing and dimensions

(a) Picture of LUCY II pressure setup
∆P
Flow
40 40 40 40 40 40 20 10 25 10 35 32 32 32 32 32 32 32

(b) Pressure tap location in LUCY II setup
Probe
Figure 3.4: Convergence test sectionSmall
for pressure
scale measurements in small scale.
z-y view Vertical

1 Dynamics-CETIM
von Karman Institute for Fluid
1
Horizontal 1 0
0
x-y view -6D +6D
Flow
y
z x
Figure 3.5: Test section for optical probe measurements in small scale.
surface area ratio tested is: σ=0.64 and 1.56 for contraction and expansion geometries
accordingly. The volume flow rate of water is ranging from a minimum of 1 l/s to a
maximum of 4.7 l/s. In Table 3.3 the test matrix for pressure measurements, optical
probe measurements and visualization is presented; the different measurement positions
are specified. Upstream diameter is denoted by d1 and downstream by D2 .
The flow rate of water studied during optical probe measurements is Q=2.5 l/s. The
volumetric qualities are 6, 9 and 14 %. Acquisition time for each measurement point is
69
Figure 3: Working principle of the optical pro
Void fraction-bubble size and velocity-s
The flow rate of water studied during the
measurements is Q=2.5 l/s. The volumetric qu
and 14 %. The acquisition time for each meas
tacq =1 minute; this value was found sufficient after acquiring measurements for different
is tacq=1 minute; this value was found su
acquisition times between 30 s and 3 minutes. The profiles are structured with 16 points for different acqu
acquiring measurements
of measurement for the upstream diameter and 20 points for the downstream between 30(2secmm andstep
3 minutes. The profiles
with 16
for both cases). The error in the positioning of the probe due to the transversal mechanism points of measurement for the upstr
and 20 points for the downstream (2 mm
moving the probe is estimated to ±0.2 mm. However, there is an important uncertainty
cases). The in positioning of the pro
error in the
the determination of2:theTest
Figure section-measurement
location positions
of the probe caused and risk
by the flowof breaking the mechanism
transversal probe nearmoving the probe is e
orientation
the wall. Therefore, a mismatch of the two type of profiles extracted0.2has mm. However,
been there is an important unce
observed
determination of the location of the probe caus
in some casesTable
(horizontal-vertical).
1: Test matrix of breaking the probe near the wall. Therefor
of the two type of profiles extrac
TEST Pressure Probe Visualization observed (horizontal-vertical).
Q [l/s] 1-4.3 2.5 2 - 4.7 The void fraction is defined
β [-] 0-30 6, 9, 14 1 - 40
T
Upstream d1 [mm] 40 32 32 α = gas
Downstream D2 [mm] 32 40 40
Tacq
Gas injection position -20d1 -36d1 -36d1
where Tgas is the total time for
-6d1 -1d1 passing in front of the probe and
Measurement positions Every 1d1 - Singularity acquisition time.
+6d1 +1d1 A typical horizontal profile for the
volumetric qualities upstream and
Results and Discussion the divergence in Figure 4 is shown. Hence, th
Table 3.3: Test matrix for pressure, optical probe measurements andthe flowsingularity and the volumetric quality
visualization.
A dual optical probe, manufactured in the French company fraction distribution is concluded. The shape
RBI,associated
is used to acquire significantly changes in the downstream part
The accuracy to theinformation
flow rate about the void fraction,
measurements is: 0.5% ofis
the full scale
concentrated
(FS)in the center of the tub
mainly
measurement bubble
for a flowsize and
metervelocity.
of typeIn Figure
COPA.3 the
Theworking principle
precision of the pressure
upstreamtransducers
part the profileis seems more unifo
of the probe is briefly explained. The probe is made out of
equal to 0.25% of the FS (Validyne). Finally, the uncertainty
Sapphire with two 30 microns tips. The latter ones are small relatedmaximum
to the values
void noticed
fraction at the two edges o
experiments isenough
estimatedto be to
able to measure small
a maximum bubbles which should
of 16%. left and right of the pipe. This phenomenon b
however be bigger than 30 microns. A distance of 0.91 mm important for higher volumetric qualities. Ad
separates the two tips. The laser beam when the probe is in void fraction becomes higher with increasin
after the cone. The latter is possibly due to th
3.1.2 Large air isscale
reflected at the tip of the probe and the signal is
of bubbles that are forming pockets of air in
acquired through an acquisition box while when the probe is
in water the beam is refracted inside water. Thus, the time the tube.
In this section, progressive convergence and divergence geometries of different opening
angles are alsoforconsidered.
which the probe meets a bubble
Comparison or case
to the water of
is recorded
sudden andexpansion and contraction is Horizontal profiles
a time step diagram is extracted as it is illustrated in Figure 1
carried out. Two
3. fluids are used; air and water. The volumetric quality of air varies from 0-
30 % with bubbly flow the main flow pattern observed. Three surface area 0.8
ratios; σ=0.43,
Sapphire probe
0.65 and 2.34 are tested. The opening angles for the case of progressive singularities are 5,
Laser beam
Water 4 4
8, 9 and 15 degrees. Reynolds number (Re) of the liquid is ranging from 0.6 8·10 to 23·10 .
z/D [-]
Internal pipe diameters are 41 mm (DN40), 62.7 mm (DN65) and 78 mm (DN80).

The gas injectors for the large scale apparatus (DN40 and DN65) are0.4illustrated in Fig.
Air
3.6. The DN40 injector has 3 rows of metallic tubes and 17 holes of0.20.5 mm diameter
(Fig. 3.6(a)) and the DN65 injector 4 rows and a total of 40 half-millimeter holes (Fig.
Bubble
3.6(b)). 0
0 5 10 15 20
∆T ∆T Void fraction [%]
Voltage
3.1.2.1 Pressure Figure 4: Influence of the singularity and t
quality on the void fraction distribution (horizo
Two tips
A schematic of the horizontal air-water flow large scale facility is shown in Fig. 3.7.
with aat maximum flow rate of 65 m3 /h is In
A centrifugal pump (1)measuring sucking
Figure 5,water
verticalfrom
void fraction profiles are p
two different points
a reservoir and is controlled with a frequency inverter. Time During the experiments,for
radial position an different
air β before and a
enlargement. A stratification of the flow is not
release valve (11) connected to the tank is kept continuously open to the atmosphere to
avoid bubbles entering the circuit. A by pass valve (12) is used to prevent facility from
70 3
(a) Small gas injector DN40 (b) Large gas injector DN65
Figure 3.6: Gas injectors used in large scale facility.
water hammer phenomenon. A temperature sensor is mounted in the reservoir, to monitor

the temperature for each measurement. Two electronic flow meters are used to measure
the water flow rate (2 and 3); their maximum capacity 12 m3 /h (3) and 32 m3 /h (2),
respectively. In the case of the desired maximum flow rate, which is 40 m3 /h, the two
flow meters are used in series. A bourdon tube pressure gauge (4) is placed upstream in
the pipe to know the wall static pressure relative to atmosphere. This indication helped
to prevent excessive pressure that could lead to a breaking of the test section, made of
PMMA. Moreover, the pressure has to be high enough to allow the necessary purging
of the pressure transducers. Therefore, the absolute pressure is held constant at around
200 kPa. The setup has an upstream calming section (5) consisting in stainless steel
pipe length of 50 diameters (50d). That assures a fully developed flow after the bend.
Close to the test section, the injection of air is performed through a gas injector (6) as
indicated in Fig. 3.7. A regulation valve (7) controls the air that is supplied from a
compressor. The air flow rate is measured by a calibrated mass flow meter (8). The
design and positioning of the air injection device are such that uniform bubbly flow is
produced at the inlet of the test section. It is found that the most suitable distance for the
air injection is 20 pipe diameters upstream the singularity. After the test section, a heat
exchanger (9) is placed for maintaining the temperature constant at around 21 ◦ C during
the experiments. A draining valve is also located at the bottom of the reservoir. Finally, a
pressure regulation valve (10) controls the pressure of the system.
A detailed view of the test section is presented in Fig. 3.8. The case of a DN 40/65
(σ=0.43) divergent section with an opening angle of 8 ◦ is exemplified. At each section
of measurement, four pressure taps are placed with an angle of 45 ◦ between them as
shown in Fig. 3.8 (up left). Thus, any three dimensionality of the flow could be identified
from pressure measurement. The four taps are named as A, B, C and D according to the
schematic. Figure 3.8 (up right) depicts an overview of the test section. The setup is built
in PMMA to allow optical access. Pressure taps are placed along the tube at several points
as is shown in Fig. 3.8 (down). The distance between pressure holes is normally equal to
1 tube diameter but becomes smaller when approaching the singularity. The pressure taps
are also more dense inside and downstream the singularity. This allows better tracking
of the flow behavior in the singularity. Pressure distribution is measured upstream and
71
Figures for thesis
72 Exact position of injection?? 3 Experimental techniques-facilities

Pressure
Large scale
11
Inverter T
1 Pump
2 Big electronic water flow meter
Water tank 3 Small electronic water flow meter

4 Bourdon tube pressure gauge
1 12 5 Calming length
3 2
Water discharge 6 Air injector
9 7 Regulation valve
8 Electronic air mass flow meter
6
P 9 Heat exchanger
4 5 10 Pressure regulation valve
Test section
11 Air release valve
7 12 By pass valve
8 T Temperature measurement
Compressed air
Figure 3.7: Experimental large scale AGATHE facility.
downstream the singularity.
3.1.2.2 Test conditions large scale

Differential pressure transducers of type Rosemount are used for the pressure measure-
ments. The uncertainty associated to the pressure transducers varies from a minimum
of 0.35% to a maximum of 0.75, depending on the range of the measurement (100-20%
of the scale of the range respectively). To obtain the best accuracy possible, 4 different
pressure transducers are selected:
1. Calibrated at 0-1.6 kPa
2. Calibrated at 0-4 kPa
Every transducer is used in range that gives the best accuracy in all the covered con-
ditions. Prior to the measurements, predictions of regular pressure drop are performed by
means of Blasius (1913) and Colebrook and White (1937) formulas for single-phase and
Lockhart and Martinelli (1949) for two-phase flow. Thus, this ∆P estimation allows the
selection of the appropriate pressure transducers for each test. Additionally, for the pre-
diction of the singular pressure change in single-phase, the coefficients given by Idel’Cik
(1986) are used. The uncertainty related to the flow rate measurements varies from a mini-
mum of 0.5 % to a maximum of 1.10%. The temperature variation during the experiments
is of the order of ± 4 ◦ C with an average value of 21 ◦ C. Although a heat exchanger is used
for reducing this deviation, a small fluctuation of the temperature could not be avoided.
A discrepancy of ±5 ◦ C will change ρ and ν by 0.1 and 11 % respectively. Therefore, a
72
3.2 Characterization of facilities for SRV 73
correction of the liquid density and viscosity with temperature is performed. Sampling
frequency of the measurements is fsamp =2 Hz and acquisition time for each measurement
point is tacq =1 minute with the aim of assuring a more accurate average. In some cases
(for sudden and progressive enlargement of σ=0.65), a higher fluctuation of the signal is
observed hence, in this occurrence an acquisition time of 2 minutes is chosen. CETIM
Large scale
Large scale Pressure
A B
4 pressure taps
(45° angle between them)
D C
Aluminum table
Figure 3.8: Test section (AGATHE).
3.1.3 Summarizing tests and flow conditions

A summary of all tests in small and large scale is presented in Table 3.4.
Flow conditions of the experimental campaigns are listed in Tables 3.5 and 3.6. Table
3.5 presents the test conditions for the pressure measurements and Table 3.6 for optical
probe tests and visualization. It should be pointed out that the ReL1 number of the liquid
is based on the upstream pipe diameter d1 . For the comparison between single and two-
phase flow, ReL1 is kept constant. This is obtained by adjusting the water flow rate when
increasing the air to reach a higher volumetric quality β. Consequently, we can assume
that the total mass flux is constant since the mass of the air compared to water is negligible.
3.2 Characterization of facilities for SRV

In this part of the project, a safety relief valve of type 1 1/2" G 3" is tested under single-
phase air, water and two-phase air-water flow. A transparent (made of PMMA) model
73
Table 3.4: Summarizing matrix of different test cases studied.

Singularity d1 [mm] D2 [mm] σ [-] L s [mm] L/d1 [-] α [ ◦]
Contraction 40 32 1.56 25 1.6 9
Contraction 62.7 41 2.34 41 1.53 15
Expansion 32 40 0.64 25 0.78 9
Expansion 41 62.7 0.43 123.8 3 5
Expansion 41 62.7 0.43 75 1.83 8
Expansion 41 62.7 0.43 41 1 15
Expansion 62.7 78 0.65 52.9 0.8 8
Sudden contraction 62.7 41 2.34 - - 90
Sudden expansion 41 62.7 0.43 - - 90
Sudden expansion 62.7 78 0.65 - - 90
Table 3.5: Upstream conditions for pressure measurements.

d1 [m] Fluid Q̄ [l/s] U [m/s] β [%] Ḡ [kg/m2 s] ReL1 ·104 Flow regime
Water 1 0.8 794 3.56 Turbulent Min
Water 4.3 3.42 3414 15.3 Turbulent Max
0.04 0-30
Air 0.017 0.013 0.016 0.00354 Laminar Min
Air 0.39 0.31 0.36 0.0832 Laminar Max
Water 2.3 1.8 1750 8 Turbulent Min
Water 7 5.4 5300 23 Turbulent Max
0.041 0-30
Air 0.4 0.3 0.38 0.09 Laminar Min
Air 2.8 2.2 2.73 0.58 Turbulent Max
Water 6 1.9 1950 13 Turbulent Min
Water 10.5 3.4 3400 23.5 Turbulent Max
0.0627 0-25
Air 0.4 0.1 0.15 0.05 Laminar Min
Table 3.6: Upstream conditions for optical probe measurements and flow visualization.
d1 [m] Fluid Q̄ [l/s] U [m/s] β [%] Ḡ [kg/m2 s] ReL1 ·104 Flow regime
Water 2 2.5 2500 9 Turbulent Min
Water 4.7 5.8 5850 20 Turbulent Max
0.032 1-40
Air 0.017 0.02 0.03 0.005 Laminar Min
of the valve is constructed (according to the plan of the industrial SRV) and investigated.
Both the model and the true valve are tested under different flow conditions (compressible-
incompressible and two-phase flow) with varying parameters; the stagnation pressure,
valve opening and volumetric quality. Depending on the flow conditions, cavitation may
appear and critical regime can be reached for highest pressures. In most cases tested, local
cavitation phenomena are observed.
74
3.2.1 SRV pressure study

An experimental closed loop (for transparent SRV) is built and the pressure is measured
at different locations along the valve and the disk. The force applied on the disk can
then be calculated through the pressure measured on three pressure taps along the disk.
Measurement of the flow rates of water and air allow determining the discharge rate for
each valve lift.
3.2.1.1 SRV transparent facility
In the assembled experimental apparatus, the flow rates of air and water and the static
pressure at several locations are measured. A schematic of the horizontal flow configu-
ration with its components is shown in Fig. 3.9. Water is pumped by a centrifugal pump
(1) of maximum capacity of 65 m3 /h and variable speed control. During the experiments,
a by-pass valve (12) and air release valve (11) are used. Furthermore, temperature sensor
is placed inside the water reservoir to check temperature variation. Two electronic flow
meters (2 and 3) are used to measure the whole range of the desired water flow rate; one
of maximum capacity 12 m3 /h (3) and one of 32 m3 /h (2). In the case of the maximum
desired flow rate, which is ' 20 m3 /h, the two flow meters are used in series.
A Bourdon Tube pressure gauge (4) is placed upstream in the pipe in order to obtain
the wall static pressure relative to atmosphere. This indication helped to prevent excessive
pressure that could lead to damaging of the test section, made of PMMA. The test section
(5) includes a calming reservoir and air is supplied through a gas injector (6). Regulation
valve (7) controls the air that is supplied from a compressor. An electronic mass flow
meter (8) measures the air flow rate. Heat exchanger (9) is placed downstream the test
section to keep a constant temperature during the experiments. A drain valve is located
at the bottom of the reservoir. Finally, pressure regulation valve (10) is used to create
pressure drop, and as a result to control the pressure of the system.
The number of orifices of the gas injector DN40 is increased to 31 and 0.5 mm diam-
eter each to reach higher air flow rate (Fig. 3.10).
The safety valve model is placed on top of a small reservoir, 100 l, as shown in Fig.
3.11 (left) to simulate the industrial use of the valve. The pressure is measured at different
locations along the valve and the disk. The flow is discharged back to the large reservoir
and the loop is closed. Differential pressure is measured between a reference pressure
taken just downstream the air injection and each pressure tap in the model. The absolute
pressure is measured as well at the reference point and it is fixed at the desired value.
The spring of the valve is removed and it is replaced with a system that allows mod-
ifying and fixing the valve opening with an accuracy of 0.1 mm, as it is depicted in Fig.
3.12. The valve employed during the measurements is a model that was manufactured ac-
cording to the plan of the 1 1/2" G 3" safety relief valve. In Fig. 3.11 (right) the industrial
valve and the modified transparent model are compared.
The main core, shown in Fig. 3.13 (left), and the inlet-outlet are of the same di-
mensions (Φ 22 and DN80). Different upstream nozzle length is chosen for transparent
valve with 10 diameters upstream nozzle length while for industrial valve the nozzle has
a length of 5 diameters. Hence, both long and short nozzle type safety valve can be tested
and compared against discharged mass flux predictions from two-phase flow models. Fur-
75
Test section Valve-pressure
Large scale
11
Inverter 1 Pump
T
2 Big electronic water flow meter
3 Small electronic water flow meter
Water tank
4 Bourdon tube pressure gauge
1 12 5 Test section-calming reservoir
3 2 6 Water discharge 6 Air injector
7 Regulation valve
9 8 Electronic air mass flow meter
5 9 Heat exchanger
P
4 7 Calming 10 Pressure regulation valve
reservoir 11 Air release valve
12 By pass valve
7
8 T Temperature measurement
Compressed air
Figure 3.9: Experimental setup to study SRV model (AGATHE II).
Figure 3.10: Gas injector DN40 (with more holes).
thermore, the shape of the core is simplified with less curved surfaces for PMMA model
in order to facilitate manufacturing. However, no influence of pressure and flow behavior
due to this discrepancy downstream the valve is expected. The valve disk in the trans-
parent model has the shape of a cylinder of the same diameter as the disk on industrial
valve. Therefore, in order to calculate the equivalent pressure, this geometrical difference
should be taken into account. The force applied on the disk can be calculated through the
pressure measured on three pressure taps along its radius as demonstrated in Fig. 3.13
(right). The formula used is given by Eq. 3.1:
ZR
F= 2πrP (r) dr, (3.1)
0
76
Safety relief valve DN80
DN 80
To reservoir
DN40
Air injection DN 80
Φ 22
Flow
Φ 22
Calming reservoir 100 l
DN 40
DN 40
Figures for thesis CETIM

Figure 3.11: Schematic of the test section AGATHE II (left) and comparison of industrial
safety valve with the transparent model Valve-pressure
(right).
Valve opening Transparent

adjustment valve
Pressure
taps
Pref
Air
injection
Calming
reservoir
Figure 3.12: Picture of the test section AGATHE II and its different components.
where F the flowforce on the valve disk in N, and R the radius of the valve disk in
mm.
In Table 3.7 the exact value of the position of each pressure tap is indicated.
Table 3.7: Position of different pressure taps along the radius of the valve disk.
r [mm] 0 7 14.35 23.5
P (r) P0 P1 P2 P3
77
Figure 3.13: Transparent valve body (left) valve disk with three pressure taps (right).
where P(r) the pressure along the valve disk radius in MPa.
Finally, in Fig. 3.14(a) the positions of the front pressure taps are indicated. Pressure
is also measured on the valve in order to identify any three-dimensional movement of the
flow (Fig. 3.14(b)).
Detailed drawing of the transparent SRV test section is given in Appendix B.
3.2.1.2 Industrial SRV facility
The advantage of studying the industrial metallic valve is that it can handle much higher
pressure than PMMA. Therefore, similar investigations were carried out in industrial SRV
setup under compressible, incompressible and two-phase flow conditions. The main fea-
tures and parameters studied in each facility are briefly presented in the next paragraphs.
The disadvantages of the metallic installations were the lack of optical access upstream,
downstream and inside the valve and that only set and back pressures of the valve could
be measured. For comparison purposes, force in the valve disk is measured with a force
sensor mounted on top of the valve stem by a specially designed system. More details are
given in Appendix B.
Incompressible The experimental installation for studying water flow in industrial SRV
is shown in Fig. 3.15. A calming reservoir is connected to HP pump (up to 250 m3 /h and
7.8 MPa) through an admission valve (controlling the flow rate) which in turn is linked
to a long pipe leading to the SRV to be tested. The pressure in the free surface of the
reservoir is fixed by compressed air (or Nitrogen under pressure) that is supplied on top
of the reservoir as indicated in the sketch.
The flow rate is measured with an electronic flow meter just after the calming reservoir
connection. A discharge valve is operating as a by-pass to obtain the Best Efficiency
Point (BEP) of the pump. The tested fluid is water of industrial quality under ambient
temperature. A displacement sensor is placed above the valve stem to measure the valve
lift when dynamic (with spring) tests are performed. This installation permits determining
the opening pressure, closing pressure and discharge coefficient Kd .
78
Figures for thesis CETIM
Valve-pressure
Front view
Disk 1
Disk 3 Disk 2
Downstream 2
Φ74
Corps 1
Nozzle 3
DN80
Nozzle 2
Downstream far
Nozzle 1
Downstream 4 CETIM
Figures for thesis
Valve-pressure
(a) Front view of transparent SRV with pressure taps
Side view Up view
Corps 3 Corps 2
Downstream 3
Downstream 1
(b) Side and up view of transparent SRV with pressure taps
Figure 3.14: Transparent SRV and position of pressure taps.
Additionally, the flowforce can be obtained with a force sensor when appropriate mod-
ification of the valve is achieved. A more detailed description of the different components
and certified procedures followed during the tests is given in Corbin et al. (2009a).
Compressible A schematic of the experimental installation for compressible fluid in-

vestigation (taken from Chabane et al. (2009)) is shown in Fig. 3.16. A primary reservoir
containing compressed air under 20 MPa supplies the required air to the secondary smaller
reservoir of 2.5 m3 , PN 40. Four flow meters of Coriolis type are connected to the buffer
reservoir and a series of solenoid valves that allow choosing the appropriate flow meter
control the air flow rate. The manifold of flow meters is connected to a SRV set to 4 MPa
to protect the system. The equipment allows SRV testing of DN50, DN65 and DN100.
79
Capteur de
Displacement
déplacement
sensor
Soupape
SRV
SRV tese
tototest
So upape
Safety de sûreté
valve Compressed
A ir co mpriméair
M o teur
Variable à vitesse
speed
motorvariable M
HP pump
Level
Détecteur
HP
HP pump
HPpump
pump detector
de niveau
P o mpe haute pressio
3h-1) n
(78 bars, 250 m -1))
(78
78 (78
78 Bar,
Bar,
Bar,b, 250
250 m33hh-1
250m3h)
m
Vanne d'admissio
Admission valven
B allo n de Flow meter

Débitmètre
Calming n
tranquilisatio
Vanne de valve
Discharge décharge reservoir
Figure 3.15: Experimental setup for SRV water study (taken from Corbin et al. (2009a)).
In the facility, pressure of 4 MPa can be achieved and mass flow rate up to 13 kg/s with
precisions of 0.25% and 1% respectively. A more detailed description of the experimental
apparatus can be found in Corbin et al. (2009a).
3ULPDU\UHVHUYRLU
6DIHW\UHOLHIYDOYH 6ROHQRLGYDOYH
6DIHW\UHOLHIYDOYH
7HPSHUDWXUH
XQGHUWHVW
3UHVVXUH
)ORZ UDWHVHOHFWLRQ
Shut-off
9DQQHGH 6ROHQRLGYDOYHV
valves
VHFWLRQQHPHQW
&RQWUROYDOYHV
(PHUJHQF\YDOYHV
&RQWURO3UHVVXUH )ORZ PHWHUV
Figure 3.16: Experimental installation for compressible fluid SRV study (taken from Cha-
bane et al. (2009)).
Two-phase flow Experimental campaign was also carried out in metallic configuration
under two-phase flow. Similar conditions to the results obtained in transparent valve are
tested in order to verify the experimental findings principally concerning the flowforce
calculations. The apparatus is the same for water industrial SRV study, presented in Fig.
3.15. The difference consisted in the ability of the test section to measure the flowforce
with a force sensor under static conditions (no spring in SRV) and air injection to create
80
a uniform two-phase flow mixture upstream of the valve. A picture of the SRV, the flow
force measurement system and the gas injection is shown in Fig. 3.17.
Force
sensor
Safety
Relief
Valve
Gas
injector
DN80
Figure 3.17: Modified industrial water SRV facility for flowforce measurements in two-
phase flow.
The gas injector used in DN80 has 5 rows with total of 63 holes of 0.5 mm diameter
each and is depicted in Fig. 3.18.
Figure 3.18: Gas injector DN80 for industrial SRV facility.
81
3.2.1.3 Test conditions SRV

An overview of the test conditions of the SRV study is given in Table 3.8. The operating
fluids are air and water and a mixture of the two; different volumetric qualities β up to 50%
of air are examined. P1 is the set pressure at the inlet and L the opening of the valve. The
experimental program is carried out both in industrial valve and in the transparent model
(single-phase water and two-phase air-water flow). The seat adjustment (position of the
nozzle ring) is a varying parameter for the industrial valve; for the transparent valve it is
set at a fixed position (15 notches). This position corresponds to zero vertical distance
between the valve seat and adjustment ring as it is indicated in Fig. 3.19. The relative
distance between valve seat and ring is shown in Fig 3.19(b); the highest the position of
¾ Cette position du siège correspond
the ring the lowest the notches number. Finally, by measuring this distance, it was found
that for à
the15
tested valve
crans sureach notch unitmétallique
la soupape corresponds to approximately 55µm displacement
of the adjustment ring.
0 notches
0 15 notches
20 notches
(a) Position of 15 notches (b) Positions of 0 and 20 notches
Figure 3.19: Relative position of valve seat and adjustment ring.
The reference pressure P1 is kept constant while adjusting the other parameters. For
transparent model five different set pressures are tested between 0.1-0.3 MPa. Maximum
allowable pressure was 0.3 MPaG due to the limited resistance of PMMA.
The experimental instrumentation is the same used for singularities study in large
scale setup described in §3.1.2.1. Hence, the same uncertainties and acquisition settings
are valid for this study. The two phenomena occurring for certain conditions are cavitation
and locally “chocked” flow, thus critical regime. Cavitation happens when vapor bubbles
are formed in water due to the decrease of the water pressure below its vapor pressure.
The critical regime is reached in the smallest passage through the valve when locally the
velocity of the fluid becomes sonic. For the case of two-phase flow, this can happen when
the speed of sound drops to a very low value by increasing void fraction (demonstrated
in Fig. 2.13(b)). Hence, the flow rate remains constant by a further decrease of the
downstream pressure. As it can be seen in test matrix, for the transparent valve, due to
the restrictions of the maximum pressure, the critical regime could not be reached.
For the pressure measurements, Rosemount differential pressure transducers are used.
82
constant by a further decrease of the downstream pressure. As
it can be seen in test matrix, for the transparent valve, due to • Flashing
3.2 Characterization
the restrictions of the maximumof facilities
pressure, thefor SRVregime
critical 83
could not be obtained.
x1vg1 C pT1 P1
ω= +
Transparent valve Metallic valve v1k v1
Conditions
Water Two-phase Water Air Two-phase
P1 [bar] 1-3 1.5-3 0.5-11 0.5-11 1.5-9 (2)
L [mm] 0.3-7.3 1-7.3 0.5-7.5 0.5-7.5 1.2-7.2 • Non-
β [%] 0 2-50 0 100 0.5-51 flashing
Seat adjustment
15 15 5-20 5-15 15
[notches] x1vg1
ω=
Cavitation/non- Cavitation& Non- Non- v1k
Cavitation Cavitation - Cavitation
cavitation non-cavitation cavitation cavitation
(3)
Critical/subcritical Subcritical Subcritical Critical Sub-critical Critical Critical Sub-critical
The next step
of the method is to determine the critical flow conditions. The
Table 2: Test matrix critical (or choked flow) is reached when the flow rate is
Table 3.8: Test matrix for SRV study.
For the pressure measurements, Rosemount maximum and a further decrease of the downstream pressure
differential pressure transducers are used. The uncertainty will not change the discharge rate.
associated to the pressure transducers varies from a minimum
The uncertainty associated to the pressure transducers• varies
of 0.35% to a maximum of 0.75, depending on the range of the Critical from flowa minimum of 0.35%
tomeasurement
a maximum of 0.75,
(100-20% depending
of the scale of the on
rangethe range of the measurement (100-20% of the scale
respectively).
To obtain the best accuracy possible, 4 different pressure P1
of the range respectively). To obtain the best
transducers are selected:
accuracy possible,P2 < ( nfour
crit ⋅ P1 )
different
, = ncrit
pressure
G (4)
transducers are selected: v 1ω
Calibrated at 0-16 mbar (0-0.23 psi)
Calibrated atat
0-800 mbar (0-11.6 psi) • Subcritical
• Calibrated 0-16 mbar
Each transducer is used in the scale that will give the smallest −2 ⎡⎣ω ln ( n2 ) + (ω − 1)(1 − n2 ) ⎤⎦ P1 (5)
• Calibrated at 0-400 mbarof the desired conditions P2 ≥ ( ncrit ⋅ P1 ) , G =
uncertainty. As a result, all the range ⎛1 ⎞ v1
is covered. ω ⎜ − 1⎟ + 1
⎝ n2 ⎠
• Calibrated at 0-800 mbar
PREDICTION MODELS For the determination of the critical pressure ratio ncrit two
As already at
• Calibrated mentioned,
0-3200thembarmost common model to cases can be discriminated:
predict the mass flux through nozzles and safety valves is the
Homogeneous Equilibrium Model (HEM). The so-called -If ω > 2
Each transducer is operated in the scale that will give the smallest uncertainty. As a
result, all the range of the desired conditions is covered.
4 Copyright © 20xx by ASME
3.2.2 SRV optical probe and visualization study (LUCY III)
Same instrumentation as in LUCY II is used in the modified facility for SRV optical
probe measurements and flow visualization named LUCY III. The experimental setup is
presented in Fig. 3.20. A higher capacity centrifugal pump (25 m3 /h at 0.4 MPa absolute
pressure) is utilized to obtain the desired pressure in the calming reservoir. In Appendix
B the drawing of the test section is illustrated (Fig. B.3).
The test section with the exact positioning of the probe placement for horizontal and
vertical profiles is demonstrated in Fig. 3.21. One position is chosen upstream the valve,
which corresponds to the nozzle diameter of the industrial valve, at 3 diameters (Φ22)
distance from the block of the valve and also one measurement position is also placed
downstream of the valve 1 diameter (DN80) distance. Air is injected with the gas injector
shown in Fig. 3.10.
Flow conditions for visualization are chosen according to observations from pressure
results. Cavitation photos and films are acquired in selected conditions to obtain a qual-
itative aspect of the phenomena. A synopsis of the optical probe and flow visualization
tests is demonstrated in Table 3.9, where d1 , D2 the upstream and downstream diameters
respectively and P1 the upstream pressure.
83
Figures for thesis
84 Valve-probe
3 Experimental techniques-facilities
Large scale
8
Inverter
2 2 3
Water tank
1 Pump
1 2 Regulation valve
3 Electronic water flow meter
Water discharge
4 4 Air injector
5 5 Calming reservoir-test section
Test section 6 Electronic air flow meter (20 l/min)

7 Rotameter (1.5 l/s)
8 Air release
2 2
7 6
Compressed air
Figure 3.20: Transparent SRV setup for optical probe measurements and flow visualiza-
Valve-probe
tion (LUCY III).
Large scale
Vertical
Downstream
Horizontal
Upstream
Horizontal
Vertical
Air
Injector
Figure 3.21: Test section of LUCY III with detailed view of the optical probe positioning.
3.3 Measurement techniques

For the measurement of water flow rate, an electromagnetic flow meter (or mag meter)
is chosen. For gas flow rate, electronic device of the same type and in certain cases
a rotameter are operated. For measuring the pressure drop, variable reluctance pressure
transducers of type Validyne and Rosemount are utilized. Furthermore, local void fraction
information is achieved through optical probe mounted in the pipe. Dual optical probe is
used to extract the local bubble velocity and size. In the next paragraphs, more details for
84
3.3 Measurement techniques 85
Table 3.9: Optical probe and flow visualization tests in AGATHE II setup.
Optical probe Visualization
Test
Two-phase Cavitation Two-phase
P1 [MPa] 0.15 0.1-0.3 0.1-0.3
Positions 1d1 /1D2 Valve disk Upstream-Body-
Downstream
Positions probe Horizontal/Vertical - -
β [%] 2, 9 - 1.5-tracer
L [mm] 7.3 2/4.5/7.3 2/4.5/7.3
each measurement technique are given.
3.3.1 Flow meters

In the present investigation, only single-phase flow rate (separately for air and water) is
measured. Operation of flow meters in two-phase flow is possible and was studied by
several authors but is out of the scope of this study. For the measurement of the flow
rate differential pressure, positive displacement and electromagnetic flow meters can be
employed. The latter one is preferred since it is the most commonly used apparatus for
this purpose.
The working principle of flow measurement by electromagnetic flow meter is based
on Faraday’s electromagnetic induction rule: when a conductor crosses a magnetic field,
voltage is generated at both ends of conductor. When a magnetic field is added at right
angles to the electrically conductive fluid, the electromotive force that is proportional to
the flow is caused at right angles to a magnetic field and the electric conductive fluid. The
direction of the electromotive force is based on Fleming’s right-hand rule, as shown in
Fig. 3.22 (right), and is given by the following formula:
E = kBDv
where:
E: Electromotive force
k: Constant
B: Magnetic flux density
D: Diameter of pipeline
v: Velocity of fluid
Therefore, for cylindrical pipe, the volume flow rate will be given by :
πDE
Q = πD2 /4 ⇒ Q =
4kB
To apply this principle to flow measurement with a magnetic flow meter, it is nec-
essary first to state that the fluid being measured must be electrically conductive for the
Faraday principle to apply. As applied to the design of magnetic flow meters, Faraday’s
85
Law indicates that signal voltage (E) is dependent on the average liquid velocity (V), the
magnetic field strength (B) and the length of the conductor (D) (which in this instance is
the distance between the electrodes). A schematic of an for
Figures electromagnetic
thesis flow meter and
its working principle is presented in Fig. 3.22.
Figure 3.22: Working principle of electromagnetic flow meter (from Copa-XE DE43F).
3.3.2 Pressure measurements

The usual type of pressure transducer is characterized by the existence of two internal vol-
umes, separated by the sensing membrane, and connected to the outside by two ports (Arts
et al. (2001)). The variable reluctance sensor is strain-based, wherein a magnetic circuit
is formed, and the parameter input causes mechanical deflection of the spring member as
a function of pressure, force, or acceleration. To provide a static output capability, vari-
able reluctance sensors require an oscillator and demodulator system internally limiting
operational temperatures from -40 ◦ C to +120 ◦ C. The spring member is comprised of
magnetic, high-permeability material and is centrally located between two coils as shown
in Fig. 3.23.
The transducer, in the case of Validyne type, is symmetrical. For differential pres-
sure transducers (Fig. 3.23), both ports are measurement ports whether, for instance, for
“gauge pressure” measurements, one of the two ports is the measurement port and the
other is open to ambient.
One of the most important characteristics of pressure transducers is the frequency
response. A high frequency response characterizes a good pressure transducer. We have to
point out that it is not correct to speak about the frequency response of a specific pressure
transducer. Rather one has to speak about the time or frequency response of a measuring
system. A measuring system consists of the transducer plus the possible fittings, plus the
pressure tap or the internal flow characteristics of a pressure probe. Estimation of the
frequency response of the transducers used can be found in Kourakos et al. (2007).
86
Figure 3.23: Variable reluctance differential pressure transducer(Validyne (2005)).
3.3.3 Optical probe

The subject of determining the void fraction is of crucial importance in two-phase flow
since the flow structure and interaction of the two phases can be identified. Additionally,
this information is very useful in pressure drop calculations. The most commonly used
method in academic and industrial applications is the optical probe technique. The advan-
tage of this procedure is that it provides local profiles of the void fraction and therefore
permits creating a 3D map of flow structure in the pipe by combining acquired profiles.
The disadvantage is that it is an intrusive method and hence can create perturbation in the
flow. When using dual probe, bubble velocity and size can be likewise resolved.
In the following sections, the principle of the optical probe system is analyzed and
main features and limitations are mentioned.
3.3.3.1 Void fraction

The working principle of optical probe is briefly explained in Fig. 3.24; it is based on the
difference of the refractive indices of water and air. The material of the probe has similar
refractive index with water. As the tip is cut with a certain angle (typically 45 ◦ ) when the
light comes across the tip surface depending on the surrounding medium in which it is in
contact, it can be either refracted or totally reflected, being this defined by the Snell’s Law
(published in Descartes (1673)) for refraction (Eq. 3.2).
n0 sinθ0 = n1 sinθ1 (3.2)

Therefore, when the probe is surrounded by water, the beam is refracted inside the
flow (zero voltage signal) while when the probe is in gaseous environment, the laser beam
is reflected at the tip of the probe originating a higher voltage signal (typically higher than
5 volts) and the signal is acquired through laser receiver to an acquisition box. Thus, the
87
acquired through an acquisition box while when the probe is pockets of ai
88 in water the beam is refracted inside water. Thus,
3 Experimental the time
techniques-facilities
for which the probe meets a bubble or water is recorded and 1
a time step diagram is extracted as it is illustrated in Figure
time for which the probe meets a bubble or water is recorded and a time step diagram is
3. as illustrated in Fig. 3.24.
extracted 0.8
Sapphire probe
Laser beam 0.6
Water
z/D [-]
0.4
Air 0.2
Bubble 0
0
∆T ∆T
Voltage Figure 4: In
quality on th
Two tips
In Figure 5, v
measuring at
radial positio
two different points Time enlargement.
of the pipe
Figure 3.24: Working principle of optical probe. (6 %). The
Figure 3: Working principle of the optical probe downstream
The time duration of the high voltage recording is proportional to residence time of
the air bubble in the tip. Consequently, in order to measure the local void fraction, the
total residence time of the bubbles for a certain position is divided by the total acquisition
time: 3
T gas
α=
T acq
where Tgas is the total time for which gas is passing in front of the probe and Tacq the
total acquisition time. This definition of the local void fraction is similar to the one given
in Eq. 2.6.
3.3.3.2 Bubble velocity and bubble diameter

During the present study, a dual optical probe, manufactured in the French company RBI,
is used to acquire information about the void fraction, bubble size and velocity. The probe
is made out of a resistant material, in this case of Sapphire, with two 30 microns tips. The
latter ones are small enough to be able to measure small bubbles which should however be
bigger than 30 microns. A distance of 0.91 mm separates the two tips. A schematic of the
two tips and their positioning in the flow according to the manufacturer is shown in Fig.
3.25(a) (taken from ISO (2008)). However, since the setup did not allow this placement
88
of the two tips, another configuration, shown in Fig. 3.25(b), is chosen. The distance
separating the two tips is indicated in the figure. The probe will record the transient time
of thethe
bubble to cross
specified between
maximum bubblethe first and the second tip (named time of flight T f light ) and
number.
since the distance between them is known (∆x), the velocity can calculated as follows:
Both criteria have the same level of priority. This means that, in reality, the acquisition ends as soon as one of
∆x
these two stopping criteria is satisfied. Vb = . (3.3)
T f light
5.1.4 where
Setting∆xthe
is the distance
geometry between the reference and the secondary sensors and T f light
characteristics
If is anareaverage
you using a transit timeprobe
two-sensor required for the
and intend bubblesthetovelocity
to measure move of from the reference
interfaces, you must sensor
enter thetotip
spacing, which defines
the secondary sensor. the distance between the tips of your two-sensor probe, in the direction of the flow. This
information is provided by the manufacturer of the sensor.
Flow
Tip
spacing
(a) Dual optical

Figure probe tips characteristics
7: Geometrical positioned in flow
of the(taken
probefrom
ISO (2008)).
The tip spacing must be given in millimetres (see tag 6 in Figure 6).
You can identify the current probe with a string composed Acquisition
of, at most, 20 characters (see tag 7 in Figure 6).
boxappears on the interface, thus enabling an easy
When processing data provided by a given probe, this legend
identification of the used sensor.
5.1.5 Selecting to perform a single acquisition Tflight («rising») Tflight («falling»)

When the push-button labelled Multiple acquisitions is off (see tag 8 in Figure 6), a single acquisition
can be performed at a time.
5.1.6 Validating or cancelling the settings Ch 0

To use the information appearing on this panel as the current parameters, you must validate your capture and
close the panel. This operation can be achieved by pressing the button labelled VALIDATE (see tag 9 in
d
Figure 6). The parameters will be stored in Windows Registry and used for later acquisitions.
If you want to cancel your capture and close the panel, press the button labelled CANCEL (see tag 10 in
nair=1 nwater=1.33
Figure 6). The information appearing on the panelChwill1be discarded and the current acquisition parameters
stored in Windows Registry will not be modified.
(b) Dual optical probe placed in flow for the current study
5.2 Multiple acquisition parameters

Figure 3.25: Probe tips inside flow.
Instead of achieving a single acquisition, a sequence of multiple acquisitions can be automatically performed.
In theWhen
interface dedicated
the tips to acquisition toparameters,
are perpendicular when the the
the flow direction, push-button labelled
downstream tip is in the
Multiple
acquisitions is high lightened (see tag 1 in Figure 8), a secondary panel labelled CAMPAIGN
PARAMETERS is accessible (see tag 2 in Figure 8). 89
12
wake of the upstream one. Thus, the void fraction values should be extracted from the
value given from the first tip. Moreover, it should be pointed out that at high velocities,
a gas pocket is trapped in this wake and creates perturbation in the second tip. This can
cause difference in void fraction and in bubble number between the two channels, and this
is the reason for the failure of cross correlation in some cases.
First tip meeting the flow should be the shortest in order to allow the piercing of the
bubble from the primary tip to the secondary. Hence, a trajectory inside the bubble is
“captured”. This leads to a chord rather than a diameter. This chord length is calculated
as:
lg = Vb T gas ,
The histogram of this set of values gives the chord length distribution function h(l).
Based on a theory describing the stereology properties of a population of spheres (Gunder-
sen and Jensen (1983)), a statistical processing provides the diameter distribution function
g(l) through the following differential equation:
" #
1 d (h (l))
g (l) = h (l) − l .
2 d (l)
When using the software provided with the probe (ISO), the following quantities are
defined:
Nb
fint = ,
T acq
4 fint
Ai = ,
Vb
6α
D sm = ,
Ai
where fint the bubble interference frequency in s−1 , Nb the total number of bubbles
passing from the probe, Ai the area interfacial concentration in m−1 and D sm the Sauter
mean diameter in mm.
The subsequent assumptions for the previous formulas are made:
• All the bubbles are spherical

• All the bubbles have the same diameter
• All the bubbles are moving with the same velocity
The bubble velocity can also be evaluated by cross-correlating the time signals respec-
tively collected on the reference and the secondary tip. If the cross-correlation exhibits a
peak value for a time delay then knowing the two tips spacing the bubble velocity can be
calculated.
More details on the optical probe measurement technique can be found in Dehaeck
et al. (2003) and François et al. (2003).
90

Experimental facilities for the study of geometrical accidents and safety relief valve are
presented in this chapter. Pressure measurements are performed in small (LUCY II) and
large scale (AGATHE) testing smooth and sudden contraction and expansion sections.
During the measurements, single-phase water flow is initially analyzed and afterwards
uniform bubbly flow is created with a gas injector and pressure evolution is measured and
compared with single-phase flow. In all cases the volume flow rate of water is constant
and air is added in the pipe. Pressure taps are placed before, along and after the geomet-
rical discontinuity in order to draw conclusions on its influence on pressure head loss.
Volumetric qualities up to 40% are reached with maximum water flow rate of ' 4.5 l/s.
Additionally, visualization is carried out in selected convergence geometry. For the same
case, optical probe measurements are performed at 6 diameters upstream and downstream
the geometry to determine local void fraction, bubble diameter and velocity information.
Safety relief valve of type API 1 1/2" G 3" is selected and tested in air, water and two-
phase flow. Transparent model of this valve with same inlet and outlet dimensions and
small discrepancies from the core and disk of industrial valve is constructed and studied
(AGATHE II). However, important difference between the two valves is found in upstream
nozzle whose length for transparent model is twice as long as in industrial valve. Same
transparent block is used to measure upstream and downstream profiles (void fraction and
bubble velocity) with optical probe and to perform flow visualization in the valve (LUCY
III).
At the last part of this chapter, the several measurement techniques employed during
the study are reported and details concerning their working principle, their limitations
and assumptions are provided. Electronic flow meters are used to measure the water and
air flow rate and in some occasions (for higher flow rates in LUCY II) rotameters have
been utilized. Variable reluctance differential pressure transducers were measuring the
pressure evolution at several locations of the facilities. Finally, void fraction, bubble size
and velocity measurements are accomplished with the aid of a dual optical probe placed
in perpendicular direction to the flow. Ratio of the total residence time of the bubbles
with total acquisition time will give an average value of the local void fraction at each
position in the duct. Moreover, the distance between two tips allows determining the
bubble velocity, defined as distance between the two tips divided by the transient time of
the bubble to cross from primary to secondary tip. Statistical analysis of measurements
can also provide information on bubble chord and diameter making the assumptions that
all the bubbles are moving with the same velocity, they have the same diameter and are
spherical.
91
Part III
Geometrical singularities
4 Expansion singularities
In this chapter experimental results obtained in various expansion configurations are pre-
sented. The chapter is divided in three main parts; pressure, visualization and optical
probe results obtained in singularities. Additionally, comparison with literature models
is presented and new pressure drop correlations proposed for each geometry have been
established and are reported in the following sections.
4.1 Pressure measurements

One of the main objectives of the study is the determination of the pressure distribu-
tion along different geometrical discontinuities. Figure 4.1 indicates the methodology
followed during the measurements in divergence section and the technique to determine
singular single and two-phase pressure change.
In the plot, it is shown that following a normal decrease upstream the expansion,
the pressure will increase to a maximum value inside the divergence (due to larger area)
and will start decreasing after a certain length in a regular way. We can split the whole
phenomenon in three regions; the upstream fully developed flow, the transitional region
with a recirculation zone and the downstream fully developed flow. The length of the
transitional region varies with ReL1 , σ, and the type of singularity.
In all tests, measurement of regular and singular static pressure changes are referred
to the pressure measured at ' 10d upstream the singularity. The singular pressure change
∆P sing is determined by extrapolating the regular static pressure drop from the start of
the singularity up to reattachment point. Since the points downstream are not enough
to obtain fully established flow and extrapolate until the beginning of expansion, regular
pressure drop is calculated by means of Blasius (1913) and Colebrook and White (1937)
formulas for single-phase and the Lockhart and Martinelli (1949) model for two-phase
flow.
The final singular
pressure change is computed by a simple summation of these three
terms (∆Preg−meas + ∆P sing−meas + ∆Preg−calc )1 . The reattachment length L/d (in terms
of the upstream diameter) is determined as the location of the maximum recovery pressure
from the start of the singularity (Kourakos et al. (2009)).
All the results are expressed in difference of total pressure in order to quantify pressure
head loss caused by the divergence. Hence, for a common reference and comparison
reasons, a dimensionless pressure drop coefficient is established. We should also point
out that in all two-phase experimental campaign, the liquid volume flow rate was kept
1
Reg, sing, calc and meas denote regular, singular, calculated and measured pressure change respectively
96 4 Expansion singularities
ΔPregular (calculated) ΔP [mbar]

ΔP [kPa]
Preference
Pmax ΔP=0
ΔPsingular
ΔPSINGULAR ΔPsingular
(measured)
-FINAL (measured)
Preference
ΔPregular
Effect of vena contr
ΔP=0 (measured)
Singularity Axial position [z/d]

S
z/d=0
Flow α Flow
Fully developed flow Transitional flowFully developed flow Fully developed flow
Inlet Outlet Inlet
Figure 4.1: Explanation of the way to determine the singular pressure change in expansion
geometry. von Karman Institute for Fluid Dynamics-Centre Technique des Industries Méca
constant while adjusting the air mass flow rate to obtain the desirable volumetric quality.
4.1.1 Determination of pressure drop coefficient ζ

The procedure to determine the pressure drop coefficient (called ζ from now on) with a
general definition for all geometries is explained. The adopted methodology consists of
the following steps:
1. Singular static pressure change ∆P sing

st is measured by differential pressure transduc-
ers.
2. Dynamic upstream and downstream total pressures are calculated and singular total
pressure drop ∆Ptot
sing
is determined.
3. Pressure drop coefficient ζ is deduced by dividing with the liquid dynamic pressure
Pdyn
L based on the smallest diameter.
The set of equations used to calculate pressure drop coefficient ζ are reported hereun-
der:
96
4.1 Pressure measurements 97
1 ∗ 2
∆Ptot
sing
= −∆P sing
st + ρ U t2 − U t1 ,
2
(4.1)
2
sing
−∆Ptot
ζ= 1 (4.2)
ρ U2
2 L L1
where ρ∗ the mixture density calculated with Eq. 2.18. Subscript 1 corresponds to the
upstream section for expansion and downstream section for contraction whilst subscript 2
indicated vice versa notation. Symbol L stands for the liquid phase and t for the summa-
tion of air and water. The singular static pressure change ∆P sing
st is determined according
to plots 4.1 and 5.1 for expansion and contraction accordingly. Finally, U is the average
phase velocity. For the case of single-phase flow, ρ∗ and Ut are replaced by liquid density
and average velocity respectively. All units are expressed in SI system.
Summarizing tables with all pressure drop results obtained in expansion geometries
are given in Appendix §K.1.
4.1.2 Sudden expansion

Figure 4.2 provides an example of single-phase measurement obtained in sudden expan-
sion of σ=0.43, for ReL1 =8.4·104 . The static pressure measured at the four peripheral taps
on the tube sections, close to the singularity (points A, B, C and D) is plotted along with
their average (point M). We can observe a small difference between the 4 points when
approaching the singularity which can be explained by a slight misalignement of the two
assembled pieces. The standard deviation of the signal is also plotted for each pressure
measurement point as an uncertainty bar.
The prediction from Blasius (1913) formula for regular loss is deduced whilst the sin-
gular pressure change is estimated with coefficients proposed by Idel’Cik (1986). Hence,
continuous black line represents the theoretical estimation or pressure distribution. Exper-
imental regular and singular pressure changes are in good agreement with the predicted
values (Kourakos et al. (2010c)).
For the same geometry and for ReL1 =1.82·105 , the two-phase static pressure change
along the pipe and singularity is plotted in Fig. 4.3. The single-phase result is also
drawn on the same graph. Two-phase experimental data are compared with prediction of
the singular pressure change in axisymmetric abrupt area expansion geometry obtained
from the models of Jannsen and Kervinen (1966), designated with red dashed line, and
Chisholm (1969) assigned with the blue dashed with dots line. These correlations were
analytically presented in §2.3.2.1. Both models rely on the assumption of a homogeneous
flow.
The plot in Fig. 4.3 shows that Jannsen and Kervinen (1966) model fits satisfactorily
with experimental results while Chisholm (1969) model slightly overestimates the pres-
sure change which is in accordance with the remarks reported by Velasco (1975). In the
region between 1 and 2 diameters downstream the singularity for both single and two-
phase flow, a small fluctuation of the pressure line is observed denoting the presence of
flow detachment.
To better emphasize the effect of two-phase flow, we define the dimensionless static
pressure change ΦLst as follows:
97
Cylindre 40/65
Re8.38E4-Elargissement brusque
0.6
Sudden enargement σ=0.43
0.5 Single-phase-ReL1=8.4·104
0.4
A B
0.3
∆P [kPa]
0.2
D C
0.1
Idel'cik [1986]
0.0 Point M
-0.1 Point A
-0.2 Point B
Point C
-0.3
Point D
-0.4
-10 -8 -6 -4 -2 0 2 4 6 8 10 12 14
x/d [-]
Figure 4.2: Single-phase static pressure change versus axial position for sudden enlarge-
ment of σ=0.43 and for ReL1 =8.4 ·105 . Comparison of experimental results with Idel’Cik
(1986) calculation. Cylindre 40/65
Re1.82E5-Elargissement brusque-17% air
4.5
4.0 5
Two-phase-17%air-ReL1=1.82·10
3.5
3.0
2.5 A B
2.0
∆P [kPa]
1.5 D C
1.0 SP-Experimental
0.5 L-M&Chisholm [1969]
L-M&Jannsen[1966]
0.0
Point M
-0.5 Point A
-1.0 Point B
-1.5 Point C
Point D
-2.0
-10 -8 -6 -4 -2 0 2 4 6 8 10 12 14
x/d [-]
Figure 4.3: Two-phase static pressure change versus axial position for sudden enlarge-
ment of σ=0.43 and for ReL1 =1.82·105 -comparison with experimental single-phase and
with models of Jannsen and Kervinen (1966) and Chisholm (1969).
∆PTS ing
P
ΦLst = , (4.3)
∆PSS ing
P
where ∆PTS ing

P
is the singular two-phase static pressure change as explained in Fig.
4.1 and ∆PS ing the corresponding single-phase. Figure 4.4 displays the evolution of the
SP
experimental ΦL versus volumetric quality at ReL1 =2.0·105 . The data are compared to
98
the model of Jannsen and Kervinen (1966) and Chisholm (1969), respectively. As it
was previously mentioned, Jannsen and Kervinen (1966) correlation agrees better than
Cylindre
Chisholm (1969) correlation with experimental 40/65
results.
Re2.0E5-DPsing
1.40
1.35 ReL1=2.0E5
1.30
1.25
1.20
[-]
1.15
L
st
Φ
1.10
1.05
Experimental
1.00
Chisholm (1969)
0.95
Jannsen (1966)
0.90
0 2 4 6 8 10 12 14 16 18 20
Volumetric quality β [%]
Figure 4.4: Dimensionless singular pressure change ΦL versus volumetric quality. Com-
parison with models of Jannsen and Kervinen (1966) and Chisholm (1969).
Comparison of static singular pressure change for various β and ReL1 is shown in Fig.
4.5 in the two different sudden expansion configurations (DN40/65 and DN65/80). For
the same liquid Reynolds number the velocity is much lower for the case of expansion
of σ=0.65, this leads to a much lower dynamic pressure and therefore to lower singular
pressure change than the enlargement of σ=0.43. Additionally, one can observe a more
emphasized influence of the geometrical discontinuity and presence of two-phase flow on
pressure drop for the second case.
4.1.3 Comparison smooth-sudden expansion

Three different opening angles of smooth expansion of σ=0.43 are tested; 5, 8 and 15 ◦
(half of the total opening as shown in Fig. 4.1). A possible detachment of the flow is
tested. According to Comolet (1963) the critical angle is ' 7 ◦ (full opening). Thus,
no flow separation is expected in this case. From comparison of the pressure measured
at different points A-B-C-D we can conclude that a slight discrepancy between the four
positions can be observed in single-phase water flow for certain conditions (ReL1 =1.8·105
for 8 ◦ opening angle), a much insignificant appearance of this occurrence can be noted
for 15 ◦ and no difference is detected for 5 ◦ angle. This phenomenon can be explained by
a slight misalignment of the three assembled parts (shown in Fig. 3.8).
Compared to sudden expansion, a progressive enlargement will create for the same
flow conditions, less pressure loss and accordingly will exhibit a higher pressure recovery
as depicted in Fig. 4.6. The up left part shows a single-phase ∆P diagram along abrupt
area expansion and divergent of angles 5, 8 and 15 ◦ , of surface area ratio σ=0.43 and at
ReL1 =1.8·105 . In the up right part of the figure, the same type of plot is built for β=17%
of air.
99
Cylindre 40/65 and 65/80
10
Single-phase Sudden enargement σ=0.43
9 Air 3%
Air 5%
8 Air 9%
Air 11%
7 Air 14%
Air 17%
∆Psing [kPa]
6 Air 18%
5
4 Sudden enargement σ=0.65
3
2
1
0
80000 120000 160000 200000 240000
ReL1 [-]
Figure 4.5: ∆P sing for several ReL1 from 0-18% of air for sudden enlargement of surface
areas σ=0.43 and σ=0.65.
From the plot, one can observe that for single-phase the pressure drops 17% passing
from divergent section of 5 ◦ , to 15 ◦ and 29% from 5 ◦ to sudden expansion and, for two-
phase flow, 11% and 21% respectively. Additionally, we can notice that all the curves in
Fig. 4.6 up right are shifted to higher x/D, meaning that the flow becomes fully developed
further downstream the singularity and thus the reattachment zone is longer in two-phase
flow. In the case of sudden enlargement, contrary to smooth divergence, the pressure, in
the vicinity of the singularity, before starting to increase, slightly decreases at 1d and then
rapidly increases again at 2d upstream of the singularity. This is due to the presence of
the secondary motion captured by the first two pressure taps. This behavior has been also
reported by Aloui et al. (1999).
Similar comments concerning the influence of the geometry and the two-phase flow
pattern can be made for the plots presented in Fig. 4.6 (bottom) for sudden enlargement
and smooth enlargement of angle 8 ◦ and σ=0.65. Single-phase water and two-phase flow
of 5% volumetric quality of air are exemplified. Additionally, contrary to abrupt area
explanation of lower σ, the phenomenon described in the previous paragraph does not
appear; possible expanation is that it occured closer to singularity and it was not captured
by the first pressure taps downstream the expansion. Finally, the pressure recovery length
is longer in two-phase flow compared to water flow; 5-7 upstream diameters for single-
phase and 7-9 diameters for two-phase flow.
A practical correlation to establish the length over diameter ratio (LS /d) corresponding
to the strongest pressure recovery for a diffuser has been proposed by Comolet (1963).
This formula (Eq. 4.4) is function of the opening angle α0 and is valid for L s /d≤13 to 15:
LS 0.22
= (4.4)
d sin2 (2α0 )
100
where L s the length of the diffuser. For the present case this correlation cannot be
applied since the ratios tested are an order of magnitude smaller (Table 3.4).
5 5
Singularity σ=0.43 Singularity σ=0.43
4 Single-phase-ReL1=1.8·10
5
4 Two-phase 17 % air-ReL1=1.8·105
3 3
∆P [kPa]
∆P [kPa]
2 2
1 1
Sudden enlargement Sudden enlargement
0 0
Divergent-angle 5° Divergent-angle 5°
-1 Divergent-angle 8° -1 Divergent-angle 8°
-2 -2
-10 -8 -6 -4 -2 0 2 4 6 8 10 12 14 16 18 -10 -8 -6 -4 -2 0 2 4 6 8 10 12 14 16 18
x/d [-] x/d [-]

2.5 2.5
Singularity σ=0.65 Singularity σ=0.65
Single-phase-ReL1=2.3·10
5
2.0 Two-phase 5 % air-ReL1=2.3·105
2.0
1.5 1.5
∆P [kPa]
∆P [kPa]
1.0 1.0
0.5 0.5
0.0 0.0
Sudden enlargement -0.5 Sudden enlargement

-0.5
-1.0 -1.0
-8 -6 -4 -2 0 2 4 6 8 10 12 -8 -6 -4 -2 0 2 4 6 8 10 12
x/D [-] x/D [-]
Figure 4.6: Comparison of pressure distribution for different expansion geometries in

single and two-phase flow.
The variation of different angular positions A-B-C-D in the tube for two-phase flow
is investigated in Fig. 4.7. Two cases are demonstrated; sudden expansion and smooth
enlargement of 8 ◦ and σ=0.65. The position of flow recirculation eddy can be followed
through separation of the two pressure lines A-B and C-D. Assuming that there is no
change in time of the location of pressure lines, a higher pressure drop of C-D is observed
from -0.5d upstream the expansion while this is inversed gradually until 2d downstream.
Hence, PAB <PCD upstream and PAB >PCD downstream which indicates air being concen-
trated upwards before the singularity, due to buoyancy effect while a portion of liquid is
entrained in the upper part (in the case of a quasi-symmetrical recirculation eddy) of the
pipe downstream expansion. This latter has been demonstrated by means of visualization
technique by Deniz et al. (2009).
The previously presented results are verified by the following statements of Comolet
(1963): “...If the opening angle of the diffuser is very small the kinetic energy is trans-
formed slowly in pressure energy, without a lot of losses but an important length of the
diffuser is needed. Therefore, the pressure drop is mainly due to friction on the wall as
in the case of a long cylindrical pipe. If the angle is big enough, the diffuser is short
and the pressure drop due to friction is smaller while the pressure drop due to mixing
is strong. If the divergence is very open, a jet that is applied and is oscillating against
the wall is formed. Hence, the pressure drop is analogous to the one produced in sudden
enlargement”.
101
Cylindre 65/80
Re2.34E5-Elargissement brusque-6%air
2.0
5
1.5
1.0
∆P [kPa]
0.5
0.0 SP-Experimental
Point M
-0.5 Point A
Point B
-1.0 Point C
Point D
-1.5
-8 -6 -4 -2 0 2 4 6 8 10
x/D [-]
enlargement of σ=0.65.
Cone 65/80
(a) Pressure evolution for sudden
Re1.75E5-Divergent 65/80 angle 8-2%air
2.0
Enargement σ=0.65, angle 8°
5
1.5
1.0
∆P [kPa]
0.5
SP-Experimental
0.0 Point M
Point A
-0.5
Point B
-1.0 Point C
Point D
-1.5
-8 -6 -4 -2 0 2 4 6 8 10 12
x/D [-]
(b) Pressure evolution for smooth enlargement of σ=0.65 and angle 8 ◦ .
Figure 4.7: Pressure evolution along expansion geometries; the position of recirculation
eddy can be identified.
4.1.4 Summary of expansion pressure measurements

Plots of pressure drop coefficient ζ versus volumetric quality β for various ReL1 are pre-
sented in this section. The cases of sudden expansion with σ=0.43 and smooth expansion
of α=15 ◦ opening angle are depicted in Fig. 4.8 and Fig. 4.9 respectively. Increasing β
results in higher ζ for both configurations while upstream liquid Re number seems to have
a more significant impact on ζ for smooth enlargement. Thus, the presence of liquid phase
is more emphasized for smoother change of section, leading to conclusion that mixing of
two phases is better for more sharp expansions.
Similar conclusions can be drawn for the case of smooth expansion of smaller opening
angle i.e. 8 ◦ and 5 ◦ shown in Fig. 4.10. Differences of 50-70 % on ζ for the same
102
Sudden expansion σ=0.43

0.50
ReL1=9.87·104
0.45
ReL1=1.66·105
0.40 ReL1=1.21·105
0.35
ReL1=2.00·105
0.30
ReL1=2.41·105
ζ [-]
0.25
0.20
0.15
0.10
0.05
0.00
0 3 5 8 10 13 15 18 20
Figure 4.8: Pressure drop coefficient ζ versus volumetric quality for

different ReL1 in sudden expansion σ=0.43
Smooth expansion σ=0.43, angle 15°
0.50
0.45
ReL1=1.66·105
0.40
0.35 ReL1=1.21·105
ReL1=9.87·104 ReL1=2.00·105
0.30
ReL1=2.41·105
ζ [-]
0.25
0.20
0.15
0.10
0.05
0.00
0 3 5 8 10 13 15 18 20 23 25

different ReL1 in smooth expansion σ=0.43, α=15 ◦ .
volumetric quality are noticed from minimum to maximum upstream liquid Re number.
In contrast, for sharper geometries, deviations from 10 to maximum of 25% are found in
pressure drop coefficient.
In order to obtain a more clear image on the impact that two-phase flow has on pres-
sure drop, two-phase multiplier ΦL is defined as follows:
ζL
ΦL = (4.5)
ζT P
where ζL and ζT P the pressure drop coefficients of liquid and two-phase mixture re-
spectively. Therefore, the two-phase multiplier corresponds to the ratio of total pressures
drop in liquid and two-phase flow.
The aforementioned parameter is plotted for various volumetric qualities in several
expansion configurations depicted in Fig. 4.12. Increased pressure drop compared to
single-phase water flow up to 50 to 70 % is observed for the case of sudden expansion and
103
Smooth expansion σ=0.43, angle 8°

0.50
0.45
0.40
0.35
ReL1=9.87·104 ReL1=1.21·105
0.30
ReL1=1.66·105
ζ [-]
0.25
0.20 ReL1=2.00·105
0.15
ReL1=2.41·105
0.10
0.05
0.00
0 2 4 6 8 10 12 14 16

different ReL1 in smooth expansion σ=0.43,
Smooth expansion σ=0.43,α=8
◦
angle 5°.
0.50
0.45
0.40
0.35 ReL1=9.87·104
ReL1=1.21·105
0.30
ReL1=1.66·105
ζ [-]
0.25
0.20 ReL1=2.00·105
0.15
0.10 ReL1=2.41·105
0.05
0.00
0 2 4 6 8 10 12 14 16

different ReL1 in smooth expansion σ=0.43, α=5 ◦ .
smooth expansion of angle 15 ◦ respectively. Much higher can become ΦL for smoother
expansions; up to 180% for 5 ◦ opening angle. More explicitly, taking an example of
constant ReL1 =1.66·105 and β=10%, the following values are found:
• Sudden expansion σ=0.43: '20 % increase in ΦL

• Smooth expansion σ=0.43, α=15 ◦ : '18% increase in ΦL
4.1.4.1 3D comparative plots in expansion geometries

From previously presented results the dependency of pressure drop coefficient on inde-
pendent parameters volumetric quality and upstream liquid Re number was concluded.
104
Sudden expansion σ=0.43 Smooth expansion σ=0.43, angle 15°

1.60 1.80
5
ReL1=1.66·10 ReL1=1.66·105
1.50 1.70
1.60
5
1.40 4 ReL1=1.21·10
ReL1=9.87·10 1.50 ReL1=1.21·105
ReL1=9.87·104
1.30 1.40 ReL1=2.00·105
ΦL [-]
ΦL [-]
5
1.20 ReL1=2.00·10 1.30
1.10 1.20
5
ReL1=2.41·10 ReL1=2.41·105
1.10
1.00
1.00
0.90
0.90
0.80 0.80
0 3 5 8 10 13 15 18 20 0 3 5 8 10 13 15 18 20 23 25
Volumetric quality β [%] Volumetric quality β [%]
Smooth expansion σ=0.43, angle 8° Smooth expansion σ=0.43, angle 5°

2.20 3.30
ReL1=1.21·105
5
2.00 ReL1=1.21·10
ReL1=9.87·104 ReL1=1.66·10
5
ReL1=9.87·104
2.80
5
1.80 ReL1=2.00·10
ReL1=2.00·105 ReL1=1.66·10
5
1.60 2.30
ΦL [-]
ΦL [-]
5
1.40 5
ReL1=2.41·10
ReL1=2.41·10 1.80
1.20
1.30
1.00
0.80 0.80
0 2 4 6 8 10 12 14 16 0 2 4 6 8 10 12 14 16
Volumetric quality β [%] Volumetric quality β [%]
Figure 4.12: Two-phase multiplier ΦL versus β for various expansion geometries.
Therefore, 3-dimensional plots of ζ function of β and ReL1 are extracted and are pre-
sented for all expansion geometries in Fig. 4.13. A more uniform distribution of ζ over
Re number and volumetric quality is observed for sharper expansion geometries. The lat-
ter can be explained by the better mixing of two phases and the more symmetrical flow
separation occurring with increasing diffuser opening angle.
Finally, the effect of opening angle in expansion geometry of σ=0.43 for different
ReL1 in single and two-phase flow is investigated and the result is plotted in Fig. 4.14.
Pressure drop coefficient is suddenly increasing for the smallest opening angles up to
about 20 ◦ and then ζ tends to stabilize towards a constant value. Since between 15 and
90 degrees there are no points, no clear conclusion on the exact transition angle can be
drawn. Influence of two-phase flow is mainly noticed for the lowest ReL1 (water flow).
The lateral can be explained by a better mixing of flow for higher liquid flow rate and
more uniform bubbly flow compared to lowest values of ReL1 .
Similar plot of ζ=f(α) has been established by Comolet (1963) for σ=0.11 in single-
phase water flow and it is depicted in Fig. 4.15. The shape of experimental curve in Fig.
4.14 agrees with the one obtained by Comolet (1963). Since the surface area is lower,
higher velocities will result in higher ∆Pdyn and as a consequence higher ζ. A minimum
of the pressure drop coefficient is found for 3.5 ◦ which corresponds to the best pressure
recuperation of the diffuser; this value is below the ones studied and consequently does
not appear in our experimental graph.
105
Figure 4.13: 3D comparison plot of expansion singularities.
4.1.4.2 Developing length in expansion geometry
Developing length in expansion singularity is defined as the area in the vicinity of en-
largement in which the flow is still developing until the fully developed flow section (as
shown in Fig. 4.1). In the present study, the limits of this region are identified by pres-
sure measurements (area from beginning of singularity up to maximum pressure recovery
point).
In Fig. 4.16 the flow developing length after expanding section is depicted for various
opening angles and two surface area ratios; σ=0.43 and 0.65. Higher L/d is found for
lower α and towards 90 ◦ the developing length reaches asymptotically a constant value.
Hence, for smoother diffusers, longer length is needed for pressure recovery. Since more
angles have been investigated for lower σ the shape of curve is more representative for
this case. Error bars indicate the uncertainty in determining the developing length due to
distance between each pressure tap; the latter being of the order of one upstream diameter.
The parameter of developing length in abrupt area expansion geometry was studied
by Ahmed et al. (2008). A correlation taking into account surface area ratio and upstream
liquid Re number is proposed:
106
5
Expansion σ=0.43, ReL1=9.87·104 Expansion σ=0.43, ReL1=1.66·10
0.5 0.5
0.45 0.45
0.4 0.4
0.35 0.35
0.3 0.3
ζ [-]
ζ [-]
0.25 0.25
0.2 0.2
0.15 0.15 Single-phase water
0.1 Single-phase water 0.1 Two-phase 4 %
0.05 Two-phase 4 % 0.05 Two-phase 10%
Two-phase 8% Two-phase 12 %
0 0
0 20 40 60 80 100 0 20 40 60 80 100
Angle [°] Angle [°]
5 5
Expansion σ=0.43, ReL1=2.00·10 Expansion σ=0.43, ReL1=2.41·10
0.5 0.5
0.45 0.45
0.4 0.4
0.35 0.35
0.3 0.3
ζ [-]
ζ [-]
0.25 0.25
0.2 0.2
0.15 0.15
Single-phase water Single-phase water
0.1 Two-phase 5 % 0.1 Two-phase 5 %
0.05 Two-phase 9% Two-phase 10%
0.05
Two-phase 12 % Two-phase 12 %
0 0
0 20 40 60 80 100 0 20 40 60 80 100
Angle [°] Comolet Angle [°]
Figure 4.14: Effect of opening expansion angle on the pressure drop coefficient for various
upstream mass fluxes.
Figure 4.15: Pressure drop coefficient for different opening angles of a diffuser (left), flow
inside diffuser (right)-taken from Comolet (1963).
Ld
= 13.788 · Re0.11
L1 (1 − σ)
2.463
, for σ ≥0.0625 (4.6)
D
Ahmed et al. (2008) have proved high dependency of L/d over ReL1 which is probably
due to the low Re numbers investigated (working fluid oil). Lower ReL1 can lead to con-
siderably different flow regimes downstream the expansion which in turn can drastically
increase L/d. In the present study upstream liquid Re number was found not to have an
important effect on L/d. Surface area ratio and opening angle were the dominant parame-
ters. In Fig. 4.17 plot of L/d function of surface area ratio is shown for sudden expansion.
Estimation of this parameter is also made with Eq. 4.6 proposed by Ahmed et al. (2008).
107
0.0625, the turbulence intensity decreases from about fraction relaxes first, followed by the liquid velocity and
17% at z/D = 14 to about 5% for JL = 0.27 m/s and then the turbulence intensity. The developing length
JG = 0.2 108
m/s (Fig. 13c). Typically, the turbulence intensity increases with both an 4 increase
Expansion
in ReL1 singularities
and a decrease in
in the upper part of the pipe is higher than in the lower theExpansion
area ratio. With the constraint that the developing
part. The higher turbulence intensities in the upper part length approaches zero as r approaches 1, the non-dimen-
can be attributed to the bubbles
16 which are present in this sional developing length can be correlated to the upstream
region, which can increase the turbulence intensity. liquid Reynolds number and area ratio using the current
14
The void fraction, liquid velocity and turbulence inten- data as
sity profiles for the stratified wavy flow (area ratio
12 Ld 2:463
0.0625, JL = 0.011 m/s and JG = 3.75 m/s) are presented ¼ 13:788 Re0:11
L1 ð1 rÞ ð1Þ
in Fig. 14. The void fraction profiles show a clear demarca- D
10
tion between the bottom liquid and upper gas layers
L/d [-]
8
for r P 0.0625. This correlation, albeit a simplified one
(Fig. 14a). Here, the void fraction increases sharply across
that does not consider the mass quality or the upstream
the interface from zero in the liquid layer to approximately
6 flow regime, was found to fit the current data reasonably
unity in the gas region. As the flow develops, the thickness
of the liquid layer increases and
4 is consistent with the flow
visualizations (Fig. 6). The liquid velocity, as expected, 30
increases from the wall to a2 maximum value near the DN40/65, σ=0.43
gas–liquid interface (Fig. 14b). It should be noted that no DN65/80, σ=0.65
reliable measurements of the 0 liquid velocity of the 20
entrained liquid droplets in the0 gas phase can 20 be made. 40
Ld
60 80 100
The turbulence intensity decreases as the flow developsOpening angle [°]
downstream of the sudden expansion and reaches an D
approximately constant value in the fully developed region 10
(Fig. 14c).
Figure 4.16: Developing length versus opening angle for enlargement singularity.
0
3.4. Developing length downstream of expansion 0 0.5 1 1.5 2
A relative discrepancy of around 20% is noticed possibly

0.11
dueReto (1 − σ )significantly
L1 ⋅ the
2.463
dif-
The influence of the liquid mass flux on the developing
ferent
length was foundrange
to beof Re numbers
much investigated.
more significant than the InFig.
Fig.16.4.18 comparison
Comparison of current of
dataother existing
with the corre-
correlation for the
developing length downstream of a sudden expansion.
gas masslations
flux. Thefor
developing length, non-dimensionalized
developing length in sudden expansion geometry is presented.
Expansion
30 30
14
Area ratio 0.0625 Ahmed et al. [2008] Re L1 = 2000 Aloui et al. [1999]
Experimental results
12Area ratio 0.25
Area ratio 0.444 Attou and Bolle [1999]
Re L1 = 820
10 20
20 Ahmed [2008]
Ld 8 Ld
L/d [-]
Re L1 = 80
D 6 D
10 10
4
0 0
0
0.4 0.45 0.5 0.55 0.6 0.65 0.7 0 0.2 0.4 0.6 0.8 1
10 100 1000 10000
Re L1 σ [-] σ
Fig. 15. Variation of the developing length with the upstream liquid Fig. 17. Comparison of current correlation with existing correlations for
Reynolds number. the developing length downstream of a sudden expansion.
Figure 4.17: Experimentally determined Figure 4.18: Developing length in expan-
developing length in sudden expansion for sion singularity versus surface area ratio
different σ-comparison with Ahmed et al. proposed by different authors (taken from
(2008) correlation Ahmed et al. (2008))
4.2 Flow visualization in expansion singularities

The determination of the flow regime is one of the most important aspects when dealing
with two-phase flow since the structure of the flow and interaction of the two phases can
considerably alter the pressure drop. Therefore, flow regime maps have to be considered.
This campaign of visualization is performed, using a high-speed camera, in fully
transparent setup. Four different flow patterns are identified downstream of the cone;
Bubbly, Plug, Disperse and Annular flow. The flow conditions for which these regimes
108
4.2 Flow visualization in expansion singularities 109
were visualized are reported in Table 3.6. Furthermore, we should draw attention to the
fact that all flow conditions calculated refer to the upstream position. Indeed, for these
test cases, flow regime upstream the singularity corresponds to bubbly flow (Baker map)
Results paper
while downstream three additional flow patterns occur (plug, disperse and annular).
In Fig. 4.19, the four different flow patterns identified with a high-speed camera di-
rectly downstream the divergence areFlowillustrated.
regimes
Bubbly Plug
Disperse Annular
von Karman Institute for Fluid Dynamics-Centre Technique des Industries Mécaniques
Figure 4.19: Flow patterns identified downstream of the divergence geometry of α=9 ◦ 30
and σ=0.64.
Normal digital camera is also used to visualize the flow in sudden and progressive
enlargement and contraction. For the same flow conditions, i.e. ReL1 =1.8·105 and air
volumetric quality of air 7%, in all expansion geometries the corresponding picture is
shown in Fig. 4.20. For the smallest opening angles of the diffuser, a recirculation eddy
that is concentrated towards the highest part of the pipe is observed while for sharpest
expansions a more symmetrical recirculation is detected.
Flow charts
Flow regime maps are often considered in two-phase flow. A common chart is the one
proposed by Baker Baker (1954). It has been established for horizontal flows in pipes
of constant cross section. In the present study, the flow is visualized both upstream and
downstream the singularity.
For abrupt and progressive enlargement (angles 5 ◦ and 8 ◦ ) with σ=0.43 and σ=0.65,
a normal video camera is used to determine the condition for transition from bubbly flow
to other types of flow just after the singularity. The results are plotted on Baker (1954)
map and are reported in Fig. 4.21. However, since the departure from bubbly flow is
decided on visual information, the transition criterion remains rather subjective and the
results given in Figure 4.21 are only indicative.
109
Divergence 40/65, angle 5° Divergence 40/65, angle 8°
Divergence 40/65, angle 15° Sudden expansion 65/80
Conditions ReL1=1.85·105, Air=7%
Figure 4.20: Visualization in enlargement geometries for ReL1 =1.85·105 and 7% volu-
metric quality of air.
The second campaign of visualization is performed, using a high-speed camera, in

a fully transparent setup that allows better optical access (without pressure taps). Con-
sequently, distinction between flow regimes is more straightforward. In this facility, a
progressive enlargement of σ=0.64 for an opening angle of α=9 ◦ is tested. The flow
conditions for which these regimes are visualized are stated in Fig. 4.22(a). We should
draw attention to the fact that all flow conditions calculated refer to the upstream posi-
tion. Indeed, for these test cases, the flow regime upstream the singularity corresponds to
bubbly flow (Baker (1954) map) while downstream three additional flow patterns occur
(plug, disperse and annular).
In Fig. 4.22(b) a representation of the downstream flow patterns in terms of upstream
liquid and gas superficial velocities (JL and JG respectively) is depicted. The transition
lines between different flow regimes are drawn and therefore, referring to upstream flow
conditions, the departure from bubbly flow upstream to another flow pattern downstream
the divergence section can be concluded. Additionally, in the same plot, the correspond-
ing pressure drop coefficient for the equivalent flow conditions is indicated at the right
110
4.3 Optical probe measurements in divergent section 111
Results paper
100
Singularities σ=0.43
and σ=0.65
Wavy
Annular
GG1 / λ [kg.m s ]
-2 -1
10
Stratified
Slug
Bubbly
1 Sud.enl.-σ=0.43
Sud.enl.-σ=0.65
Plug Div.angle 5 σ=0.43

Div.angle 8 σ=0.43
Div. angle 8 σ=0.65
0.1
1 10 100 1000 10000 100000
-2 -1
GL1ψ [kg.m s ]
36
Figure 4.21: Modified Baker (1954) map for progressive and sudden expansion of σ=0.43
and 0.65.
ordinate of the plot.
4.3 Optical probe measurements in divergent section

Dual optical probe measurements are performed in the selected configuration; divergence
of σ=0.64 and angle 9 ◦ . Test conditions are constant water flow rate 2.5 l/s and three
different values of β; 6, 9 and 14%. Local void fraction and bubble diameter and velocity
are acquired for different flow rates and positions (Kourakos et al. (2010b)).
A typical horizontal void fraction profile for the three different volumetric qualities
upstream and downstream the divergence are presented in Fig. 4.23(a). Hence, the in-
fluence of singularity and volumetric quality on void fraction distribution is concluded.
The shape of profile significantly changes in downstream part where air is concentrated
mainly in the center of the tube while in upstream part the profile seems more uniform
with two maximum values noticed at the two edges of the pipe i.e. left and right of the
pipe. This phenomenon is more accentuated for higher volumetric qualities. Addition-
ally, void fraction becomes higher with increasing β but also after the cone. The latter is
possibly due to the coalescence of bubbles that are forming pockets of air in the center of
the tube.
In Fig. 4.23(b), vertical void fraction profiles are plotted against radial position for
different β before and after the pipe enlargement. A stratification of the flow is noticed
upstream of the pipe especially for the lowest volumetric quality (6 %). The void fraction
also becomes higher in downstream part and distribution of the void fraction seems more
111
Results paper
100
Wavy Divergence σ=0.64

Annular
Angle 9°
GG1 / λ [kg.m s ]
-2 -1
10
Bubbly
Slug
1
Stratified
Bubbly
Plug Disperse
Plug
Annular
0.1
1 10 100 1000 10000 100000
-2 -1
GL1ψ [kg.m s ]
(a) Modified Baker Baker (1954) map with the four patterns identified downstream
the singularity. 37
2.5 0.16
Divergence σ=0.64 Downstream
Angle 9° flow regime
0.14
2.0 Plug
Annular Bubbly 0.12
Disperse
Upstream JG [m/s]
Annular 0.1
1.5
Plug
ζ [-]
0.08
1.0
0.06
Disperse
0.04
0.5
0.02
Bubbly
0.0 0
2.0 3.0 4.0 5.0 6.0 7.0
Upstream JL [m/s]
(b) Flow pattern regions with transition lines for downstream section. Upstream su-
perficial velocities of water JL and air JG are indicated.
Figure 4.22: Flow map for progressive expansion of σ=0.64 and α=9 ◦ .
uniform after the singularity although still the highest amount of air is concentrated in the
upper part of the tube due to the buoyancy effect. Finally, we should point out that the void
fraction increase due to singularity is more significant for the case of highest volumetric
qualities. Thus, the influence of the pipe expansion on void fraction values seems to be
more important for the case of higher air-water mixtures.
Figure 4.24 shows the flow structure given by visualization, comparing with the results
112
Horizontal profiles
1 Horizontal profiles
-6D-β =6%
1
-6D-ββ=6%
+6D- =6%
0.8 +6D-ββ=9%
-6D- =6%
0.8 -6D-ββ=9%
+6D- =9%
+6D-ββ=14%
-6D- =9%
0.6 -6D-ββ=14%
+6D- =14%
+6D-β =14%
[-] [-]
0.6
z/Dz/D
0.4
0.4
0.2
0.2
0
0 5 10 15 20 25 30
0
0 5 10 Void fraction
15 [%] 20 25 30
Void fraction [%]
(a) Influence of singularity and volumetric quality on void fraction distribution (hori-
zontal profile).
Vertical profiles
1 Vertical profiles
1
0.8
0.8
0.6
[-] [-]
0.6
y/Dy/D
0.4 -6D-β =6%

0.4 -6D-ββ=6%
+6D- =6%
+6D-ββ=9%
-6D- =6%
0.2 -6D-ββ=9%
+6D- =9%
0.2 +6D-ββ=14%
-6D- =9%
-6D-ββ=14%
+6D- =14%
0
0 20 40 60 80 +6D-β =14%100
0 Void
0 20 40 fraction [%]60 80 100
Void fraction [%]
(b) Influence of singularity and volumetric quality on void fraction distribution (ver-
tical profiles).
Figure 4.23: Horizontal and vertical profiles upstream and downstream divergence sec-
tion.
obtained with optical probe in the upstream and downstream position for three volumetric
qualities of air; 6, 9 and 14%. For this purpose, void fraction profiles are plotted function
of the radial distance of the pipe. The accordance between qualitative and quantitative
results is satisfactory. A considerable stratification of the flow is remarked mainly above
9% of volumetric quality. In the downstream section, a formation of plugs of air is noticed
for the two highest air volume fractions.
Optical probe by means of a dual tip provides also information about the average bub-
113
Upstream Downstream
1 1
-6D-β =6% +6D-β =6%
0.8 0.8
6 % air 0.6 0.6
y/D [-]
y/D [-]
0.4 0.4
0.2 0.2
0 0
0 20 40 60 80 100 0 20 40 60 80 100
Void fraction [%] Void fraction [%]
1 1
-6D-β =9%
0.8 0.8
0.6
0.6
y/D [-]
y/D [-]
9 % air 0.4
0.4
0.2
0.2
+6D-β =9%
0
0 0 20 40 60 80 100
0 20 40 60 80 100 Void fraction [%]
Void fraction [%]
1 1
-6D-β =14%
0.8 0.8
0.6 0.6
y/D [-]
y/D [-]
14 % air
0.4 0.4
0.2 0.2
+6D-β =14%
0 0
0 20 40 60 80 100 0 20 40 60 80 100
Void fraction [%] Void fraction [%]
Figure 4.24: Flow visualization upstream and downstream divergence section-comparison

with void fraction profiles.
ble velocity, the mean Sauter diameter and the bubble size distribution. The aforemen-
tioned measured quantities will be discussed in in this paragraph. The average bubble
velocity is given by Eq. 3.3. By interpreting the results provided by the software ISO
of the probe, an average bubble diameter and velocity for the duration of acquisition for
each measurement point is extracted.
The Sauter mean diameter is the diameter of a monodispersed bubble for which the
volume-surface ratio is equal to that computed for the actual bubble. The definition of
this parameter is given by the formula:
3
d30
D32 = 2
, (4.7)
d20
where d30 is the volume mean diameter denoting the diameter of a monodispersed
bubble equivalent to the actual bubble in liquid volume and d20 is the surface mean diam-
eter representing the diameter of a monodispersed bubble equivalent to the actual bubble
in liquid surface.
In addition, from experimental campaign, the bubble size distribution in each mea-
surement position is deduced. In the graph representing the bubble distribution, one can
identify a maximum value (the peak of the curve) which corresponds to the most probable
diameter. An example of such a diagram in the upstream section, at the horizontal plane
and the middle of the pipe for β=14%, is given in Fig. 4.25. The maximum diameter,
114
called D peak , is specified. The log-normal distribution is attempted for fitting with this
case.
β =14 %-Upstream-Horizontal-z/D=0.53
1
lognormal
Dpeak=1.439 experimental
0.8
0.6
PDF [-]
0.4
0.2
0
0 2 4 6 8 10
Bubble diameter [mm]
Figure 4.25: Example of bubble size distribution diagram with log-normal fit, peak diam-
eter D peak identified.
Filtering at 300 μm
In Fig. 4.26(a), the bubble size distribution is extracted in horizontal plane for up-
stream and downstream locations and β=6%. The most probable diameter D peak is indi-
cated together with the associated void fraction at the same location. From the plot, one
can conclude that for the same position i.e. the center of the tube (z/D=0.5), the maximum
diameter remains practically constant at 1 mm. A slightly higher diameter is observed for
the right side of the pipe (z/D=0.84) for a similar local void fraction of 8% while for the
left part, smaller diameter for low void fraction (1%) is measured. Therefore, the bubble
diameter is not strongly affected from the singularity for this specific volumetric quality.
In Fig. 4.26(b), a similar chart is presented for the case of β=9% and 3 positions in
the pipe before and after the enlargement. For both cases we can observe that the largest
diameters are observed towards the center which is in accordance to the horizontal void
fraction profiles presented previously. By comparing the equivalent positions upstream
and downstream, we can detect that the cone considerably alters the distribution of the
bubble population in z-direction and this phenomenon seems stronger with increasing
volumetric quality. Moreover, at the center of the tube, D peak is two times higher in down-
stream part compared to upstream section. This can possibly be explained by coalescence
of bubbles. The distribution seems not to approach a log-normal fit for the case of down-
stream section at the center of the pipe. Contrary to lower volumetric quality, for this
occurence, the singularity seems to play a significant role in bubble size distribution.
Finally, the effect of volumetric quality at the same location (z/D=0.5) before and after
the singularity is examined in Fig. 4.27. Both in upstream and downstream parts, the
maximum D peak is found for the intermediate volumetric quality (β=9%). For 9 and 14%
115
β =6 % -6D-horizontal
Bubble population [-] Bubble population [-]

[-] [-]
200 z/D=0.53 600 Dpeak=0.625
population
200 z/D=0.53 600 Dpeak=0.625
population
400
100 Dpeak=1.004 (α=0.76%)
400
100 Dpeak=1.004 (α=0.76%) 200 Dpeak=1.004 (α
Dpeak=1.004 (α
Bubble
0 200
0 1 2 3 4 5 6 7 8
Bubble
0
0 Bubble diameter [mm] 0 2
0 1 2 3 4 5 6 7 8 0
β =6 % diameter
Bubble +6D-horizontal
[mm] 0 2
[-] [-]
Dpeak=0.558 (α=1.11%)
β =6 % +6D-horizontal
200 2000
population
z/D=0.26 D
Dpeak=0.558 (α=1.11%)
200 Dpeak=1.051 (α=8.07%) z/D=0.47 2000
population
z/D=0.26 D
100 Dpeak=1.051 (α=8.07%) z/D=0.84
z/D=0.47 1000
100 Dpeak=1.226 (α=7.54%) z/D=0.84
1000 Dpeak=1.051 (
Bubble
Dpeak=1.226 (α=7.54%)
0 0 Dpeak=1.051 (
Bubble
0 1 2 3 4 5 6 7 8 0 2
0 Bubble diameter [mm] 0
0 1 2 3 4 5 6 7 8 0 2
(a) Bubble size distribution for volumetric quality β=6%.
Dpeak=0.806 (α=0.29%)
100 1
=0.988 (α=1.14%)
D
Dpeak=0.806 peak
(α=0.29%) Dpeak=1.466 (α=1.11%)
z/D=0.2
100 Dpeak=0.988 (α=1.14%) z/D=0.53 1
Dpeak=1.466 (α=1.11%)
z/D=0.2
50 z/D=1
z/D=0.53 0.8
50 z/D=1 0.8
0 0.6
0 1 3 2 4 5 6 7 8
[-] [-]
0 Bubble diameter [mm]
1 0 2 0.6
β3=9 % +6D-horizontal
4 5 6 7 8
y/D y/D
Dpeak=0.475 (α=7.23%) Bubble diameter [mm]
=3.231 (α=11.57%) 0.4
D α=7.23%) β =9 %D+6D-horizontal
=1.231 (α=3.32%) peak z/D=0.21
400Dpeak=0.475 (peak 0.4
Dpeak=1.231 (α=3.32%) Dpeak=3.231 (α=11.57%) z/D=0.47
400 z/D=0.21
z/D=0.84 0.2
200 z/D=0.47
z/D=0.84 0.2
200
0
0 1 2 3 4 5 6 7 8 0
Bubble diameter [mm] 0 1
0
0 1 2 3 4 5 6 7 8 0
(b) Bubble size distribution [mm]quality β=9%.
for volumetric
Bubble diameter 0 1
Figure 4.26: Bubble size distribution at the horizontal plane for β=6% and 9% at upstream
and downstream positions for different locations in the pipe.
of air, the curve changes considerably after the singularity while for 6% the distribution
remains practically the same.
The effect of gas injector was investigated by Deniz et al. (2009). Two injectors of
0.5 and 1 mm diameter holes are compared in horizontal and vertical plane upstream and
downstream singularity for the same conditions (Fig. 4.28). From comparison of the void
fraction profiles, it is concluded that a small deviation is only remarked for horizontal
upstream profiles which seem slightly more disturbed. This is possibly due to interaction
between bubbles which increases with increasing bubble diameter. Therefore, since no
significant influence of this parameter is found, a gas injector of 1 mm holes is chosen for
the final tests. While the diameter of gas injector holes does not considerably affect the
116
-6D -z/D=0.5
Bubble population [-]

z/D=0.53 600 Dpeak=0.625 (α=3.85%)
β =6 %
400 β =9 %
β =14 %
200 Dpeak=1.004 (α=0.76%)
Dpeak=1.466 (α=1.11%)
7 8 0
0 2 4 6 8 10
+6D -z/D=0.5
Bubble population [-]
z/D=0.26 2000 Dpeak=2.241 (α=23.77%)

β =6 %
z/D=0.47
β =9 %
z/D=0.84
1000 β =14 %
Dpeak=3.231 (α=11.04%)
Dpeak=1.051 (α=7.09%)
7 8 0
0 2 4 6 8 10
Figure 4.27: Influence of volumetric quality on bubble size distribution at the center of
the pipe upstream and downstream the singularity.
Horizontal profiles
1
z/D=0.2 -6D-β =9%
void fraction values, no further investigation of its influence on the bubble diameter and
z/D=0.53 +6D-β =9%
z/D=1 velocity is performed. A more detailed discussion of this subject can -6D-
be found in Deniz
β =14%
0.8
et al. (2009). +6D-β =14%
Assuming that at one instant bubble and liquid have different velocities, if we apply
the momentum
0.6 equation for the bubble surrounded by liquid in the horizontal x-direction
7 8
y/D [-]
(Fig. 4.29), we obtain:

dVb dUb 1
z/D=0.21
0.4 + mb Vb = −C D ρL Ab (Ub − U L )2
mb (4.8)
dt dx 2
z/D=0.47 where mb the mass of the bubble in kg, Vb the bubble volume in m3 equal to the
z/D=0.84
volume of sphere 34 πR3 , Ub the bubble velocity in m/s, UL the liquid velocity, CD the drag
0.2
coefficient and Ab the surface of the bubble in m2 . We want to obtain the point that bubble
reaches the same velocity with liquid; no change of bubble velocity with x-direction.
7 8 0 dU
Therefore,0 the term dxb 1of Eq. 4.8 will2be equal to zero.
3 Equation is 4then rewritten 5as:
Dsm [mm]
dVb 1 πD2
= −C D ρL (Ub − U L )2mb (4.9)
dt 2 4
Applying this formula for an example of initial liquid velocity 3 m/s in 1 mm spherical
bubble and with drag coefficient CD =64/Re, we find the relaxation time of bubble ' 150
µs. This means that a bubble that will be in a liquid environment in horizontal flow will
reach the same velocity as the liquid in t=150 µs. Hence, we can conclude that since
relaxation time of the bubble is so small, the length that will be required for a bubble
to “follow” water flow is very small (0.45 mm for the current example). Thus, we can
assume that there is no slip velocity between the two phases under these conditions and
the values of bubble velocity obtained with optical probe measurements correspond to
liquid velocity as well.
117
0,2
0 with 1 mm hole diameter injector seems more disturbed. This might be since the interaction
0 1 2 3 4 5 6 the bubbles increases with increasing bubble diameter. Upstream and downstream
118 void fraciton (%)
between 4 Expansion singularities
vertical void fraction distributions for 0.5 and 1 mm hole diameter injectors are given in the
followinginjectors
Figure 53 Upstream horizontal void fraction distribution with different figures.
0.5 mm diameter injector 0.5 mm diameter injector
3l/s-29l/min 3l/s-50l/min 3l/s-61l/min

3lt/s-29l/min 3lt/s-50l/min 3lt/s-61l/min
1,2
1,2
1 1
r/d 0,8 0,8

0,6
r/d
0,6
0,4
0,4
0,2
0,2
0
0 10 20 30 40 50 60 0
0 10 20 30 40 50 60 70 80 90
void fraction (%) void fraction (%)
1 m m diam eter injector

1 mm diameter injector
3l/s-29l/min 3l/s-50l/min 3l/s-61l/min
3lt/s-29l/min 3lt/s-50l/min 3lt/s-61l/min
1,2 1,2
1 1
0,8 0,8
r/d
0,6
r/d
0,6
0,4 0,4
0,2
0,2
0
0
0 10 20 30 40 50 60 0 10 20 30 40 50 60 70 80 90
void fraction (%) void fraction (%)
Figure 54 Downstream horizontal void fraction distribution with different injectors

Figure 55 Upstream vertical void fraction distribution with different injectors
(a) Downstream horizontal void fraction distribu- (b) Upstream vertical void fraction distribution
tion with different injectors. with
As seen in Figure 55 different injectors.
and 56, distribution and the values of two measurements seems close to
each other, which means, variation in injector hole diameter has almost no effect on void
Figure 4.28: Comparison of void fraction profiles

fraction distributionobtained for two different gas injectors
in vertical direction.
(taken from Deniz et al. (2009)).
Ub
Ul
Gas injector Bubble
Figure 4.29: Bubble and liquid movement in the pipe.
In Fig. 4.30 horizontal and vertical bubble velocity profiles are extracted for β=9% and
14%. These plots point out the influence of the presence of singularity on bubble velocity;
the increase in pipe diameter results in decreasing fluid velocity and thus Ub becomes
smaller after the diffuser. However, the influence of increasing volumetric quality seems
to affect more bubble velocity in the upstream part than in the downstream and this can
be explained by a better mixing of the two phases after the expansion (air and water).
Horizontal profiles are shown in Fig. 4.30(a). The velocities distribution is nearly
symmetrical for 6% of air while for 14% in downstream pipe, acceleration is noticed on
the two sides of the pipe due to bubbles concentrated mainly near the center of the pipe.
In Fig. 4.30(b), vertical bubble velocity profiles are illustrated; in the upper part of the
118
+6D-β =9%
0.8 -6D-β =14%
0.8
+6D-β =14%
y/D [-][-]
0.6 0.6
[-]
y/D [-]
z/D
z/D
0.4 0.4
pipe, bubble velocity is lower due to buoyancy which pushes the bubbles up and therefore
coalescence occurs. As a result, smaller 0.2bubbles
VKI will remain at the bottom part and will
0.2
Résultats-Logiciel
move with aISO
higher velocity due to blockage0
effect. 0 0 1 2 3 4 5 0 1 2 3
Dsm [mm] Vb [m/s
Horizontal profiles Vertical profiles

1
-6D-β =9% 1 -6D-β =6%
+6D-β =6%
+6D-β =9% +6D-β =6%
-6D-β =9%
-6D-β =14% -6D-β =9%
0.8 +6D-β =9% 0.8
+6D-β =14% +6D-β =9%
-6D-β =14%
-6D-β =14%
0.6
+6D-β =14%
0.6 +6D-β =14% La Vbulle
y/D [-]
[-]
y/D [-]
• Concentra
z/D
0.4 0.4
• « Blocage
0.2 0.2
VKI
0 0
4 5
Nouveau traitement 0 1 2 3
Vb [m/s]
4 5 6 0 1 2 3
Vb [m/s]
4 5 6
I)
rtical profiles
(a) Horizontal bubble velocity profiles. (b) Vertical bubble velocity profiles.
-6D-Le
β =6%temps de relaxation d’une bulle ↓↓ Æ longueur ↓
+6D-β =6%
Æ La bulle4.30:
-6D-β =9%
Figure
+6D-β =9%
« très vite » suit leofliquide
Comparison horizontal and vertical bubble velocity profiles upstream and
-6D-β =14%
downstream
V ~ 4 m/s,the
bulle
+6D-β =14% Vl,debutante
La Vbulle=3.11
singularity
estfor β=9%
m/s
plus basseand 14%.
vers le haut
9 Traitement fait par
Explication: Une partie de la section
• Concentration du gaz élevée Nouveau traitement
Explanation of increased upstream bubble Flora
velocity on the(Ingénieure
Tomasoni bottom part of the tube is
VKI)
est shown
occupéeinpar
Fig.l’air Æ ~ 30 % Æ Vbulle ↑ Le temps de rela
• 4.31(a).
« Blocage » de la section V=f(D)
Æ La bulle « très v
30% Vbulle ~ 4 m/s, V
3
Vb [m/s]
4 5
32
6
mm Explication: Un
est occupée par l’ai
Vertical profiles
1
-6D-β =14%
0.8
32 mm
0.6
y/D [-]
0.4
0.2
0.33 β=14 %,
0
0 20 40 60 80 100
Void fraction [%]
(a) Blockage of 30% of section due to concentrated (b) Diagram of bubble velocity versus chord length
bubbles on the upper part of the pipe for z/d=0.33
Amont, upstream and β=14
horizontal, z/D=0.33 β=14 %,
Figure 4.31: Explanation of blockage effect in upper part of the duct and bubble velocity
versus chord length diagram.
The starting fluid velocity is ' 3.4 m/s (Qt /A1 ) for 2.5 l/s water flow rate and β=14%.
For this case a blockage of the section ' 30 % due to stratified bubbles is found from void
fraction profile (Fig. 4.31(a)). This results in an increased fluid velocity passing from
bottom part of the tube around 4.3 m/s. Thus, this can be an explanation of the velocity
measured by the probe (4 m/s) for the same conditions.
Finally, bubble velocity versus chord length is calculated for upstream horizontal posi-
tion at z/D=0.33 β=14 % (Fig. 4.31(b)). Post-processing was performed by F. Tomasoni,
119
researcher in VKI. It is proved that most of the bubbles have the same velocity and are
following the flow at 4 m/s as shown in Fig. 4.30(a).
Evaluation of measurements with optical probe

To evaluate the reproducibility of the measurements using the optical probe technique a
group of 30 samples was taken by Fernandes et al. (2010) for constant conditions of flow
of 2 l/s and 10% of volumetric quality. These measurements were performed at 5.6D from
the singularity at the center of pipe section. The results obtained are presented in Table
4.1.
Table 4.1: Repeatability test for optical probe measurement (taken from Fernandes et al.
(2010)).
Yi αP0 [%] αP1 [%] Vc−c [ms−1 ] D sm [mm]
σ 0.57 0.68 0.06 0.08
σ/Yi [%] 11.45 10.84 2.17 5.94
In this table, α P0 stands for the local void fraction given by the secondary prong while
α P1 refers to the value given by the reference prong. The Vc−c denotes the velocity given
by cross-correlation and D sm the mean Sauter diameter. Minor variations are found for
bubble velocity and Sauter Mean Diameter and discrepancies of 10% are noticed for void
fraction values. This result can be considered satisfactory in terms of repeatability of the
measurements.
At this point we should point out that in many cases during experiments, non-logical
bubble diameter values have been registered by the probe i.e. 15 mm. This is obviously
wrong and can be explained by the limitation of probe on measurement only of spherical
bubbly flow. In many cases, due to bubbles concentrated on upper part of the pipe, bubbles
combined forming long slugs resulting in the extremely high values of bubble diameter
recorded.
4.4 Comparison with existing models

Comparative graphs of experimental results with two literature correlations are given in
Fig. 4.32 and 4.33 for one test case with conditions: ReL1 =2.0·105 and β ranging from 0-
18%. In these diagrams, the maximum deviation from the experimental data are indicated;
for model of Jannsen and Kervinen (1966) it is limited to 5% while it reaches 10% for
Chisholm (1969) model.
Overall deviations (in %) per geometry:
• Sudden expansion σ=0.43: Idel’Cik (1986):11.11, Jannsen and Kervinen (1966):13.84,

Chisholm (1969):24.21
• Sudden expansion σ=0.65: Idel’Cik (1986):20.93, Jannsen and Kervinen (1966):27.72,
Chisholm (1976):42.24
120
4.4 Comparison with existing models 121
Cylindre 40/65 Re=2,0E5
8.0
7.5 ReL1=2.0·105 5%
∆Psingular Jannsen [kPa]

7.0
10%
6.5
6.0
Single-phase
Air 3%
5.5
Air 5%
Air 9%
5.0
Air 11%
Air 14%
4.5
Air 17%
Air 18%
4.0
4.0 4.5 5.0 5.5 6.0 6.5 7.0 7.5 8.0
∆Psingular experimental [kPa]
Figure 4.32: Deviation of Jannsen and Kervinen (1966) model from

Cylindre 40/65 Re=2,0E5
experimental results.
8.0
7.5 ReL1=2.0·105 5%
∆Psingular Chisholm [kPa]
7.0
10%
6.5
6.0
Single-phase
Air 3%
5.5
Air 5%
Air 9%
5.0
Air 11%
Air 14%
4.5 Air 17%
Air 18%
4.0
4.0 4.5 5.0 5.5 6.0 6.5 7.0 7.5 8.0
∆Psingular experimental [kPa]
Figure 4.33: Deviation of Chisholm (1969) model from experimental

results.
• Smooth expansion σ=0.43, α=5 ◦ : Idel’Cik (1986):10.14
Detailed table with all deviations from literature correlations is presented in Appendix
K.1.
121
4.5 New correlations proposed

According to Comolet (1963), pressure drop coefficient depends on opening angle, sur-
face area ratio and Re number (ζ=f(α, σ, Re)) in single-phase flow. In the present case, ζ
is measured both in single- and two-phase flow and an attempt to extract empirical corre-
lations that include the influence of this parameter as well is performed. For this purpose,
data fitting by multiple linear regression is applied between dependent variable ζ and two
independent variables ReL1 and β for each expansion configuration. Correlations of the
form: ζ=f(ReL1 , β) are obtained with 99% confidence level.
The correlation obtained is of the following type:
ζ = A · ReL1 · 10−7 + B · β + Γ (4.10)

Coefficients of correlations proposed per expansion configuration are reported in Table
4.2. Average deviation and regression parameter R2 are also stated.
Table 4.2: Proposed correlation for different expansion singularities-regression parame-

ters and average deviation from experimental measurements.
Singularity A [-] B Γ R2 Deviation [%]
Expansion σ=0.65, α=90 ◦ 6.357 0.117 -0.024 0.72 8.54
Expansion σ=0.43, α=90 ◦ -2.397 0.488 0.329 0.823 3.29
Expansion σ=0.43, α=15 ◦ -1.903 0.721 0.253 0.87 5.27
Expansion σ=0.43, α=8 ◦ -8.051 1.001 0.260 0.855 9.88
Expansion σ=0.43, α=5 ◦ -12.038 1.186 0.291 0.821 18.41
Around 30 samples were used for each geometry which seems to be adequate to cor-
relate properly the data.
In order to investigate if the correlations extracted are statistically correct, the value
of P-coefficient has been checked. This coefficient expresses the possibility that there is
no linear relationship between independent variable X and dependent variable Y overall,
what is the probability that randomly selected points would result in a regression line as
far from horizontal (or further) than observed.
The value of P has been lower than 5% for all correlations except for sudden expansion
geometry of σ=0.65 for which the P-value was quite high, of the order of 25%. Therefore
non-linearity can be concluded for this case which is possibly due to very low values of ζ
and flow regime instabilities.
Deviations of experimental versus predicted values from correlation are plotted in Fig.
4.34 and Fig. 4.35 for expansion geometries of σ=0.43 and 0.65 respectively. Maximum
relative discrepancy of 30% is found with most data lying upon 10% of deviation which
can be evaluated as a reasonable agreement of proposed correlation with measured data.
An average correlation for expansion, taking into account opening angle and surface
area ratio as well, is established:
ζ = −0.772 · σ + 0.0016α − 6.2 · 10−7 ReL1 + 0.825β + 0.562 (4.11)

where α the opening angle (half angle), σ surface area ratio, ReL1 upstream liquid Re
number and β volumetric quality.
122
4.5 New correlations proposed 123
Sudden expansion σ=0.43 Smooth expansion σ=0.43, angle 15°

0.45 0.45
0.43 0.43
5% 5%
0.40
0.40
0.38
ζ predicted [-]
ζ predicted [-]
10 %
0.38 10 %
0.35
0.35 0.33
0.33 0.30
0.28
0.30
0.25
0.28 0.23
0.25 0.20
0.25 0.28 0.30 0.33 0.35 0.38 0.40 0.43 0.45 0.20 0.23 0.25 0.28 0.30 0.33 0.35 0.38 0.40 0.43 0.45
ζ measured [-] ζ measured [-]
(a) Predicted versus measured ζ for sudden expan- (b) Predicted versus measured ζ for smooth expan-
sion sion α=15 ◦
Smooth expansion σ=0.43, angle 8° Smooth expansion σ=0.43, angle 5°
0.35 0.40
0.38
0.33
10 % 0.35
0.30 15 %
0.33
0.28
ζ predicted [-]
ζ predicted [-]
0.30
0.25 20 % 0.28
0.23 0.25
30 %
0.20 0.23
0.20
0.18
0.18
0.15
0.15
0.13 0.13
0.10 0.10
0.10 0.13 0.15 0.18 0.20 0.23 0.25 0.28 0.30 0.33 0.35 0.10 0.13 0.15 0.18 0.20 0.23 0.25 0.28 0.30 0.33 0.35 0.38 0.40
ζ measured [-] ζ measured [-]
(c) Predicted versus measured ζ for smooth expan- (d) Predicted versus measured ζ for smooth expan-
sion α=8 ◦ sion α=5 ◦
Figure 4.34: Deviation of predicted-measured pressure drop coefficient for expansion

singularities of σ=0.43.
Sudden expansion σ=0.65

0.20
0.19
10 %
0.17
0.16
ζ predicted [-]
0.14 20 %
0.13
0.11
0.10
0.08
0.07
0.05
0.05 0.07 0.08 0.10 0.11 0.13 0.14 0.16 0.17 0.19 0.20
ζ measured [-]
Figure 4.35: Deviation of predicted-measured pressure drop coefficient for expansion

singularities of σ=0.65.
123
A total of 133 samples are used to fit with this correlation which are statistically
sufficient. Overall deviation from measured values is 20.41 %, R2 =0.8, multiple R=0.9
and P-value lower than 10−8 for all independent variables.
Predicted with measured ζ are compared for all expansion data in Fig. 4.36. Maxi-
mum deviation of 50% is found especially for the lowest values of pressure drop coeffi-
cient with better than 80% data fitting accuracy for most points.
Expansion geometries
0.50
α=90° σ=0.43
0.45
α=15° σ=0.43
0.40 α=8° σ=0.43 25 %
0.35 α=5° σ=0.43
ζ predicted [-]
α=90° σ=0.65
0.30
50 %
0.25
0.20
0.15
0.10
0.05
0.00
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5
ζ measured [-]
Figure 4.36: Correlation for all expansion geometries.
Finally, we should point out that all correlations rely upon specific test cases
and therefore their accuracy is limited in these conditions. If more general
formulations are needed, the validity of proposed correlations needs to be
checked against desired geometries and flow conditions.

In the first section, pressure evolution plots across several singularities are demonstrated
and singular pressure change contribution is presumed. Pressure drop coefficients are ex-
tracted and compared for each case. Flow visualization is performed in divergent section
with a high-speed camera and different flow patterns are identified for various flow con-
ditions. Resulting flow regime maps are established. Optical probe measurements are
performed in the selected case of divergence section of 9 ◦ angle and σ=0.64. The results
obtained are exemplified and void fraction profiles are plotted. The flow structure is com-
pared with flow visualization for the same conditions. The influence of section expansion
on bubble size distribution is investigated upstream and downstream the singularity. A
last fragment is dedicated in discussing the evaluation of measurements acquired with
optical probe.
124
At the second part of this chapter a comparison between literature proposed methodol-
ogy and experimental measurements is attempted. Satisfactory agreement with literature
models is found for single and two-phase flows in sudden expansion geometries. Third
part is dedicated in proposing a new pressure drop correlation extracted from experimen-
tal database for each singularity. Measured versus predicted values are compared and
validity range of these formulas is discussed.
125
5 Contraction singularities
Contrary to expansion geometry, the static pressure will always decrease in contracting
sections since dynamic pressure is increasing in downstream channel. The singularity
causes a steep drop of pressure compared to regular loss. The next sections are dedicated
to presentation of the results obtained in various convergence singularities. Pressure co-
efficients are determined and predictive experimental correlations are established. Addi-
tionally, flow visualization provides qualitative information on flow across convergence
geometry.
5.1 Pressure results

In contracting sections, a vena contracta, which is the point in a fluid stream where the
diameter of the stream is the least, is formed just after the change of section. Its effect on
pressure drop is illustrated in schematic shown in Fig. 5.1. In the same figure, an example
of measurement performed in contraction geometry and the methodology to determine
static singular pressure drop are demonstrated. The final contribution of the singularity to
pressure change will be the measured ∆P subtracting regular loss from the beginning of
convergence until the end of vena contracta; the start of fully developed flow as indicated
in the graph. The length of this region is determined by plotting the pressure evolution in
several flow conditions for each geometry and observing the shape of pressure evolution
curve. Both smooth and sudden contraction geometries are investigated. For this case,
only two opening angles are tested ' 9 and 15 ◦ since there is no particular interest in the
influence of this parameter for convergence geometry.
It is important to point out that although the vena contracta has not been
measured in current project, its presence has been observed and is “included”
in a certain way inside experimental ∆P correlations developed at the last
sections of this chapter.
5.1.1 Sudden-smooth contraction

Contraction singularity of σ=1.56 and angle 9 ◦ is initially tested. The geometry is iden-
tical to the test section shown in Fig. 3.8 with a scaling factor of 1/2 (DN40/32). The
experimental facility and flow conditions are described in §3.1.1.3. Pressure transducers
of type Validyne are used for this experimental campaign with the same acquisition time
(tacq =1 min) and sampling frequency (fsampling =2 Hz). The different membranes that cover
all the range of the pressure measurements are:
128 5 Contraction singularities
ΔP [kPa]
ular (calculated)
Preference
ΔPSINGULAR-FINAL
Pmax ΔP=0
ΔPsingular
ngular
(measured) ΔPregular (calculated)
asured)
egular
Effect of vena contracta
asured)
Axial position [z/d]

Singularity Axial position [z/d]
=0
z/d=0
Flow
Cc
onal flowFully developed flow Fully developed flow Contraction area Fully developed flow
Outlet Inlet Outlet
Figure 5.1: Explanation of the way to determine the singular pressure change in contrac-
for Fluid Dynamics-Centre Technique des Industries Mécaniques
tion geometry.
18
Additionally, numerical simulations are carried out with the commercial CFD code
Fluent. The test parameters and conditions are: 2D axisymmetric computation, realiz-
able k- turbulence model with enhanced wall treatment and second order discretization
scheme. Convergence criterion is set at 10−7 . More information of the CFD computations
are given in Bacharoudis et al. (2008).
In Fig. 5.2, experimental and numerical static pressure drop is plotted against axial
position for several ReL1 in single and two-phase flow. The pressure is decreasing in a
regular way before the singularity; the contraction creates a high pressure drop step and
then starts decreasing regularly downstream.
The flow is observed fully developed close to the singularity (at ' 2d upstream and
downstream) contrary to the case of divergence for which the reattachment length is de-
128
5.1 Pressure results 129
15
Single-phase-Exp-Re=136000
Single-phase-Exp-Re=79300
Single-phase-CFD-Re=739000
13
Two-phase-11% air-Exp-Re=95100
Two-phase-10%air-CFD-Re=66500
10
∆P [kPa]
8
5 Smooth convergence σ=1.56,

angle 9°
0
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5 0.55 0.6
L [m]
Figure 5.2: Experimental and numerical single and two-phase static pressure change ver-
sus axial position for convergence of σ=1.56 and 9 ◦ angle for several ReL1 .
tected at ' 10d. Therefore, the singular pressure change ∆P sing for convergence geometry
is determined by measuring the static pressure at equal distance upstream and downstream
the singularity (2d).
Figure 5.3 shows the variation of the experimental singular pressure drop as a function
of the volume flow rate for single-phase flow. The CFD simulation presented in the same
graph, agrees satisfactorily with data. An experimental correlation for single-phase flow
is deduced (used for the calculation of the dimensionless ΦLst ):
∆PSsing
P
= 525.05 · Q̄2 − 26.947 · Q̄ + 309.42. (5.1)
This equation indicates an offset of 309 Pa, due for this case to the sensitivity of the
transducer’s membrane used. However, this fact introduces a relatively small uncertainty
in the determination of the single-phase singular pressure drop.
A summarizing graph of all experimental and numerical results obtained for single
and two-phase flow is shown in Fig. 5.4. The results concerning the case of sudden
contraction for several σ and G are compared to experimental data for smooth contraction
(Guglielmini et al. (1997)). The results are plotted in terms of the dimensionless pressure
change ΦL , defined by Eq.4.3. In Fig. 5.4, Jannsen and Kervinen (1966) correlation for
sudden contraction (Eq. 2.35) is adapted with a correction coefficient of C=0.81 to fit
with the results (G=1990 kg/m2 s). More detailed analysis of the results obtained for this
geometry are given in Delgado-Tardáguila et al. (2008).
In the large scale facility of upstream to downstream piping DN65/40, a sudden and
progressive contraction of 15 ◦ angle are investigated. The measurements are performed
in a similar way as in the case of enlargement although less measurement points are
chosen since the developing length is much smaller. Single-phase static pressure change
in sudden and smooth contraction geometry of angle 15 ◦ and σ=2.34 are presented in
Fig. 5.5(a).
129
Cylindre 40/65
Re8.38E4-Elargissement brusque
13
Smooth convergence σ=1.56, angle 9°
Single-phase flow
10
∆Psingular [kPa]
8
3
Experiments
CFD
0
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
Qwater [l/s]
Figure 5.3: Single-phase static ∆P sing obtained experimentally and numerically for several
Q̄water in progressive convergence geometry of σ=1.56 and 9 ◦ angle.
Lockhart-Martinelli
Convergence-Comparison all results
1.55
Exp-G=1990
1.5 Exp-G=2786 G=1990
Exp-G=1990-3424
1.45
CFD-G=1300-1700
1.4 Guglielmini et al. [1997]-G and σ varying
Janssen[1966] correlation-C=0.81
1.35
ΦstL [-]
1.3
1.25
1.2
1.15
1.1
1.05 Smooth convergence σ=1.56,
angle 9°
1
0 5 10 15 20 25 30 35
Figure 5.4: Experimental and numerical dimensionless singular static pressure change
ΦLst versus volumetric quality. Comparison to literature (Guglielmini et al. (1997)) and to
adapted (C=0.81) Jannsen and Kervinen (1966) model.
The regular loss is much stronger in the downstream pipe due to increasing velocity.
The presence of vena contracta is demonstrated for sudden change of section by a very
strong drop of the pressure locally at 0.1 diameters which is stabilized in the next points at
' 0.5 diameters. For progressive convergence, the vena contracta does not appear and the
pressure gradually reaches the point of regular pressure loss. This point is attained further
130
downstream compared to sudden contraction, at 2 upstream diameters. The singular pres-

sure drop step can be easily calculated by linearly extrapolating the downstream curve up
to the beginning of the singularity. A comparison with values calculated from Idel’Cik
(1986) for the same geometry and flow conditions is performed and a small deviation of
65/40
the order of 5-8% is observed. Re2.20E5-Convergent brusque 65/40
10
Idel'cik [1986]
Ligne M
0
Ligne A
Ligne B
-10
Ligne C
Ligne D
∆P [kPa]
-20
-30
-40
A B
-50 Sudden contraction σ=2.34
Single-phase-ReL1=2.2·105 D C
-60
-10 -8 -6 -4 -2 0 2 4 6 8
x/D [-]
(a) Single-phase static pressure change versus axial position for sudden contrac-
Cone 65/40
tion.
Re2.24E5-Convergent 65/40 angle 15
5
Idel'cik [1986]
Ligne M
0
Ligne A
Ligne B
-5 Ligne C
Ligne D
∆P [kPa]
-10
A B
-15
D C
-20
-25 Smooth contraction angle 15°, σ=2.34

Single-phase-ReL1=2.24·105
-30
-10 -8 -6 -4 -2 0 2 4 6 8
x/D [-]
(b) Single-phase static pressure change versus axial position for smooth contrac-
tion.
Figure 5.5: Comparison of experimental single-phase water results with Idel’Cik (1986)
calculation for sudden and smooth contraction of σ=2.34, 15 ◦ angle and ReL1 =2.2·105 .
Same type of plot for two-phase flow is presented in Fig. 5.6. The single phase curve
is also drawn in the plot. Tested volumetric quality is equal to β=9%. A slightly stronger
pressure drop, by almost 9%, is observed for two-phase flow compared to single-phase
flow for both sudden and smooth contraction.
The formula of Comolet (1963), given in Eq. J.8, is used to calculate the predicted
pressure drop coefficient. Although this formula is established for single-phase flow, ex-
131
perimental results are in good agreement with two-phase curve which leads to the conclu-
sion that this correlation overestimates the pressure drop in single-phase flow. For the case
of sudden contraction, the impact of vena contracta is longer compared to single phase
flow since established flow is observed above 1.5 diameters downstream. Two-phase de-
veloped flow region is longer for progressive convergence compared to single-phase flow
at 3 upstream diameters after the singularity. 65/40
Re2.42E5-Contraction brusque 65/40-9%air
10
Sudden contraction σ=2.34
0
-10
∆P [kPa]
-20
-30 SP-Experimental
Point M
Point A
-40
Point B
A B
Point C
-50
Point D
L-M&Comolet [1963] D C
-60
-10 -8 -6 -4 -2 0 2 4 6 8
x/D [-]
Cone
(a) Two-phase static pressure drop versus 65/40
axial position for sudden contraction.
Re2.25E5-Convergent 65/40 angle 15-9%air
5
SP-Experimental
0 Ligne M
Ligne A
-5 Ligne B
Ligne C
Ligne D
-10
L-M&Comolet [1963]
∆P [kPa]
-15 A B
-20
D C
-25
-30 Smooth contraction angle 15°, σ=2.34

-35
-10 -8 -6 -4 -2 0 2 4 6 8
x/D [-]
(b) Two-phase static pressure drop versus axial position for smooth contraction.
Figure 5.6: Two-phase flow results in for sudden and smooth contraction of σ=2.34 and
15 ◦ angle for 9% of air and ReL1 =2.2·105 . Comparison with experimental single-phase
and Comolet (1963) formula.
5.1.2 Summary of contraction pressure results

Pressure drop coefficient (named ζ) is determined in contraction geometries. The def-
initions used were presented in §4.1.1; the pressure drop coefficient is normalized by
132
dynamic pressure in downstream channel instead of upstream in expansion. It should be

pointed out that in all results the singular pressure loss and ∆P coefficient calculated are
only due to singularity; the regular loss is subtracted.
In Fig. 5.7, the results obtained in two configurations; sudden contraction of σ=2.34
and smooth contraction of σ=1.56 and α=9 ◦ , are exemplified. Pressure drop coefficient is
plotted versus volumetric quality of air with varying ReL1 . Completely different behavior
between the two cases is observed. For smooth contraction important influence of water
phase is remarked with increasing pressure drop coefficient up to 30% (for small ReL1
drop) both in single- and two-phase flow. Additionally, air has a dominant effect on
pressure loss for lowest water flow rates (steeper slope of the curve ζ=f(β)).
In contrast, for sudden contraction, higher ReL1 have been investigated and a totally
dissimilar tendency is concluded. Variation of water flow rate seems to play no role at all
in pressure drop coefficient up to 5-6% of air while after this value, decreasing upstream
liquid Re number results in increasing ζ of about 15%. A small decrease in pressure drop
coefficient occurs for lowest water flow rates at high β and this can be explained by flow
regime transition.
Same plots are represented in dimensionless form by dividing with single-phase water
flow pressure drop coefficient for each ReL1 . Hence, two-phase multiplier ΦL is obtained
function of volumetric quality (Fig. 5.8) in order to attain conclusions about the influ-
ence of air on pressure drop. For smooth contraction, a 40% increase on pressure drop
coefficient is found for only 8% of air and highest ReL1 while for lowest amount of water,
increase of ζ up to 80% is noticed for β=10%.
It is remarkable that the two curves for lowest ReL1 (8.93·104 and 1.25·105 ) do not
tend towards unity which means that there is a very sharp increase of ζ even for small
volumetric qualities. For sudden contraction more gradual linear curve is found up to 7%
of air (30% increased ζ). Sharpest increase of two-phase multiplier occurs for higher β
(14%) reaching a value of ≈ 100%.
Summarizing tables with all pressure drop results for expansion geometries are given
in Appendix §K.2.
3D comparative plots in contraction geometries
Three dimensional plots of ReL1 -β-ζ are presented in this paragraph to better emphasize
the dependency of pressure drop coefficient on the other two parameters. In Fig. 5.9
three plots of smooth and sudden contractions are demonstrated. For sudden contraction
of σ=2.34, a progressive increase of ζ with volumetric quality is found for most condi-
tions. Furthermore, increasing ReL1 slightly decreases pressure drop coefficient especially
for the highest β, while for single-phase the latter effect becomes negligible. Smooth con-
traction of same surface area ratio results in higher fluctuation of ζ and a more accentuated
impact of ReL1 especially for lowest volumetric qualities. Moreover, significantly lower
values of ζ compared to the equivalent sudden contraction geometry are reported leading
to conclusion that sudden contraction exhibits drastically higher pressure drop compared
to smooth convergence. Finally, 3D plot of ReL1 -β-ζ for smooth contraction of σ=1.56
is presented in the same figure. We can notice large variations of pressure drop with
increasing ReL1 and β.
133
Smooth contraction σ=1.56, angle 9°
0.45
ReL1=8.93E4
0.40 ReL1=1.25E5
ReL1=1.53E5
0.35
0.30
ζ [-] 0.25
0.20
0.15
0.10
0.05
0.00
0 2 4 6 8 10 12 14 16 18
(a) Influence of volumetric quality of air on pressure drop coefficient for various ReL1
in smooth contraction of σ=1.56 and α=9 ◦ .
0.45
0.40
0.35
0.30
0.25
ζ [-]
0.20
0.15
ReL1=1.64E+05
0.10 ReL1=1.80E+05
ReL1=1.91E+05
0.05 ReL1=2.01E+05
ReL1=2.19E+05
0.00
0 2 4 6 8 10 12 14 16 18
(b) Influence of volumetric quality of air on pressure drop coefficient for various ReL1
in sudden contraction of σ=2.34.
Figure 5.7: Pressure drop coefficient in contraction singularities.
5.2 Visualization in contraction singularities

Pictures of the flow with normal digital camera in selected conditions (same as in expan-
sion geometry) for sudden and smooth contraction of σ=2.34 are taken. The formation
of contraction region (vena contracta) is illustrated for both geometries in Fig. 5.10. The
length of vena contracta is clearly restricted at the length of 3-4 pressure taps (1.6 up-
stream diameters) downstream singularity which is in accordance of pressure evolution
plots shown in Fig. 5.6(a). A zoom at the beginning of contraction shows the contracted
streamlines (bottom right of the figure). For smooth contraction smaller contracted region
134
5.3 Comparison with literature models 135

2.00
1.80
1.60
ΦL [-]
1.40
1.20
1.00 ReL1=8.93E4
ReL1=1.25E5
ReL1=1.53E5
0.80
0 2 4 6 8 10 12 14 16 18
2.00
1.80
1.60
ΦL [-]
1.40
1.20
ReL1=1.64E+05
ReL1=1.80E+05
1.00 ReL1=1.91E+05
ReL1=2.01E+05
ReL1=2.19E+05
0.80
0 2 4 6 8 10 12 14 16 18
Figure 5.8: Two-phase multiplier ΦL for various ReL1 in sudden and smooth contraction.
is visualized compared to sudden geometry.
5.3 Comparison with literature models

Comparison with literature correlations for contraction singularities is carried out in this
paragraph. For single-phase flow, coefficients are given by Idel’Cik (1986) and for two-
phase flow, the correlation of Jannsen and Kervinen (1966) is used (Eq. 2.35).
Average deviations for each configuration are calculated. Relative discrepancy of
around 35% is found for single-phase flow (Idelcik) and and 13% for two-phase flow
(Jannsen).
135
ζ [-] ζ [-]
ζ [-]
Figure 5.9: 3D plots of ReL1 -β-ζ for sudden contraction σ=2.34, smooth contraction
σ=2.34 and α=15 ◦ and smooth contraction σ=1.56 and α=9 ◦ .
Overall deviations (in %) per geometry:
• Sudden contraction σ=2.34: Idel’Cik (1986):37.16, Jannsen and Kervinen (1966):13.62

• Smooth contraction σ=1.56, α=9 ◦ : Idel’Cik (1986):45.60, Jannsen and Kervinen
(1966):-
Detailed table with deviations from literature correlations for all geometries tested at
different flow conditions is presented in Appendix K.2.
5.4 Application of ω-method in contraction geometry

In this section, an attempt to apply the ω methodology (presented in §2.4.7.2) for conver-
gence singularity is presented. This method can be used in nozzles, control valves, orifices
and safety valves. Example of sudden contraction geometry is chosen for comparison
with experimental findings and is shown in Fig. 5.11. Pressure drop coefficient is deter-
mined with ω model as a ratio of total pressure drop between upstream and downstream
section of the contraction over dynamic pressure computed based on velocity calculated
according to discharged mass flux prediction from ω method.
A satisfactory agreement is only noticed for the highest mass flux tested (G=3190
kg/m2 s). This is due to the insufficiently good mixing of the two phases for the lowest
mass fluxes which results in separated (or non-homogeneous) two-phase flow. Since ω
method assumes homogeneous flow, this factor considerably affects its accuracy. The dis-
crepancies between experimentally determined and predicted from model pressure drop
136
5.5 New correlation 137
Visualization photos selection
Sudden contraction 65/40 Convergence 65/40, angle 15°
Vena contracta
Conditions ReL1=1.85·105, Air=7%
Figure 5.10: Visualization in contraction geometries for ReL1 =1.85·105 and 7% volumet-
ric quality of air.
coefficient correspond to the strong vena contracta formed just downstream the singu-
larity as shown in Fig 5.10. Therefore, we can conclude that ω method can be applied,
except from SRV, also for simple geometries, given that the two phases are well mixed.
Useful information on the nature of the flow (i.e. vena contracta) can be provided from
application of ω-method.
5.5 New correlation

Using the same technique as for the case of expansion, empirical correlations to predict
pressure drop coefficient are developed for contraction singularities.
ζ = A · ReL1 · 10−7 + B · β + Γ (5.2)

The regression parameters A, B and Γ of correlations proposed for contraction are
shown in Table 5.1 for around 30 samples per geometry with R2 =0.9. The most satisfac-
tory fitting is found for sudden contraction with average digression of 4%.
Predicted and measured values are plotted in Fig. 5.12 for σ=2.34. Smooth contrac-
tion has led to worst fitting with maximum of 20-30% difference while sudden contraction
137
Sudden contraction-second definition
1.7
Omega-G=2390
1.6 Exp-G=2390
Omega-G=2780
1.5 Exp-G=2780
Omega-G=3190
1.4 Exp-G=3190
ζ [-] 1.3
1.2
1.1
1.0
0.9
0.8
0.0000 0.0001 0.0002 0.0003 0.0004 0.0005 0.0006 0.0007
Mass quality x [-]
Figure 5.11: Predicted with ω-method versus experimental pressure drop coefficient
against mass quality in sudden contraction of σ=2.34 .
Table 5.1: Proposed correlation for different contraction singularities-regression parame-

ters and average deviation from experimental measurements.
Singularity A [-] B Γ R2 Deviation [%]
Contraction σ=2.34, α=90 ◦ -4.122 1.264 0.270 0.951 4.18
Contraction σ=2.34, α=15 ◦ -0.846 0.112 0.025 0.802 27.57
Contraction σ=1.56, α=9 ◦ -15.196 0.659 0.326 0.914 7.65
presents very satisfactory fitting with results being among 95% from measured values.
As for expansion geometries, a lack of linear dependence of fitting variables from ζ for
smooth singularity could explain higher deviations.
Correlation for smooth contraction of σ=1.56 and 9 ◦ angle is applied in all tested
flow conditions as shown in Fig. 5.13. Reasonable agreement is concluded with 10%
dissimilarity.
An overall correlation of pressure drop coefficient in contraction function of the sev-
eral test parameters is proposed:
ζ = −0.153 · σ + 0.0033α − 9.7 · 10−7 ReL1 + 0.698β + 0.468 (5.3)

where α the opening angle (half angle), σ surface area ratio, ReL1 upstream liquid Re
number and β volumetric quality1 .
This correlation presents an average variation 19.53% from experimental results, coef-
ficient of determination R2 =0.94, multiple R=0.97, P-value is of the order of 10−20 which
is statistically significant and 93 samples acquired.
In Fig. 5.14 comparison of predicted pressure drop coefficient for all convergence test
cases illustrates important mismatch for smallest values while for highest ζ, prediction
1
Surface area ratio σ in contraction is defined as the ratio of upstream to downstream section
138

0.45
0.43
5%
0.40
0.38
10 %
ζ predicted [-]
0.35
0.33
0.30
0.28
0.25
0.23
0.20
0.18
0.15
0.15 0.18 0.20 0.23 0.25 0.28 0.30 0.33 0.35 0.38 0.40 0.43 0.45
ζ measured [-]
(a) Sudden contraction

0.035
0.030
ζ predicted [-]
0.025 20 %
0.020
40 %
0.015
0.010
0.005
0.005 0.010 0.015 0.020 0.025 0.030 0.035
ζ measured [-]
(b) Smooth contraction, α=15 ◦
Figure 5.12: Correlation for contraction σ=2.34.
inequality is below 20%. The former disagreement is possibly linked to flow regime
transition, as it is also indicated in the plot, for the contraction of 15 ◦ .

Convergence singularities are studied in this chapter; smooth and sudden geometries of
different opening angles and surface are ratios are considered. Static pressure evolution
plots are presented. Pressure decreasing regularly upstream the contraction is slightly af-
fected close to the singularity (1-2 diameters) followed by a sudden pressure drop step
and the influence of vena contracta (strong pressure peak) for maximum 3 upstream di-
ameters after the singularity. Pressure drop coefficient is deduced for each test case. For
two-phase mixtures, influence of liquid Re number is noticed above β=7% for abrupt con-
139

0.40
0.35
ζ predicted [-]
0.30 15 %
0.25
30 %
0.20
0.15
0.10
0.10 0.15 0.20 0.25 0.30 0.35 0.40
ζ measured [-]
Figure 5.13: Correlation for smooth contraction of σ=1.56 and 9 ◦ angle.

0.45
=90° =2.34
0.40 =15° =2.34
0.35 =9° =1.56
predicted [-]
0.30
Flow regime 25 %
0.25 transition
0.20
50 %
0.15
0.10
0.05
0.00
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45
measured [-]
Figure 5.14: Comparison of prediction of ζ with overall correlation for contraction ge-
ometries with experimental measurements.
traction while for smooth convergence, liquid Re number affects significantly the pressure
drop even for the smallest volumetric qualities tested. This is possibly due to the unsuf-
ficiently good mixing of the two phases for the lateral case. Flow visualization close to
the contracting section reveals the presence of vena contracta and its developing legth
noticed. Finally, experimental pressure drop correlations function of volumetric quality,
upstream liquid Re number are extracted using the same methodology as the one pre-
sented in Chapter 4.
140
Part IV
Safety Relief Valve

6 Safety Relief Valve
Objective of this chapter is to present the analysis of safety relief valve for two different
cases; industrial SRV of type API 1 1/2" G 3" (short nozzle) and transparent model of the
same valve (long nozzle). The accuracy of various existing methodologies for sizing SRV
is discussed in order to provide recommendations for their appropriate use at different
conditions. The study is focused on measuring the pressure evolution along the valve and
therefore determining the pressure drop, the discharged mass flux and discharge coeffi-
cient and hydrodynamic force applied on valve disk for various flow conditions and lift
positions. Visual observations of cavitation and injected two-phase flow are performed
in transparent model. Finally, CFD simulations in axisymmetric plane of the valve are
carried out in order to compare and validate experimental results.
6.1 Pressure and flow rate measurements

Safety relief valve (SRV) is selected for testing in this part of the study (API 1 1/2" G
3" ). Two main valves are used; an industrial one that allows high pressure testing and
an identical transparent model that can handle lower pressures but in which full optical
access is possible. In both cases the spring is removed and variation of the flow conditions
for fixed valve lift is performed. The aforementioned methodology permitted static flow
behavior investigation of the valve and “freeze” the phenomena occurring while valve
operates under compressible, incompressible and two-phase flow.
6.1.1 Industrial valve-transparent model discrepancies

Two types of installations are built and investigated; transparent valve in closed loop
circuit and industrial valve discharging in the environment. In the first case, back pressure
occurs affecting the behavior of the valve. Hence, this discrepancy is taken into account
for the flow rate and pressure drop determination. Moreover, the valve disk in transparent
valve is instrumented with 3 pressure taps which permitted to determine the flowforce
applied on the disk.
A comparison between results obtained in prototype valve and in transparent valve is
performed in order to verify that the latter one simulates satisfactorily the flow behavior
of the actual valve. Therefore, for single-phase (water) flow and a set pressure of 0.2 MPa
(relative total pressure), the volume flow rate is plotted versus valve opening for the two
cases (Fig. 6.1). A reasonable agreement given the minor geometrical discrepancies is
observed (Kourakos et al. (2010a)).
144 6 Safety Relief Valve
The flow section of the SRV is the minimum passing section. If the nozzle section
AN =πD2N /4 is larger than the lateral section A f =πDL, then fluid is discharged through
it. Therefore, as indicated from the shape of the two curves, for L≥D/4 (D=22 mm) ⇒
L≥5.5 mm the flow rate tends to stabilize. This opening corresponds to ' 70% of the full
lift.
Comparaison soupape transparente-metallique-eau 15 crans
25
Pset=0.2 MPa
20
Q [m3/h]
15
10
5
Transparent valve
Industrial valve
0
0 1 2 3 4 5 6 7 8
L [mm]
Figure 6.1: Volume flow rate of water versus valve opening for transparent and industrial
valve.
6.1.2 Single and two-phase air-water flow

Influence of the presence of air on mass flow rate versus the valve opening can be deduced
from Fig. 6.2(a) in industrial valve. The upstream set pressure of the valve P1 is equal to
0.3 MPa. From the graph, it is obvious that the highest the volumetric quality, the lowest
the mass flow rate passing through the valve and this is more emphasized with increasing
valve opening. Since gas is much lighter than liquid, for the same total volume occupied
by the two phases, lower total mass will pass through the valve creating a higher pressure
drop.
We can observe that discharged mass flow rate is linearly decreasing with volumetric
quality, i.e. for example at full lift for 20% volumetric quality mass flow rate decreases
about 21% compared to water flow. Thus, a linear regression between mass flow rate and
volumetric quality for several lifts is applied. The resulting Ṁ versus β is plotted for full
lift, L=4.5 mm and L=1 mm in Fig. 6.2(b).
6.2 Flow visualization in SRV

Flow visualization is carried out in the upstream part, downstream part and the core of
the valve. The purpose is to identify, if possible, topology of the flow downstream the
144
6.2 Flow visualization in SRV 145
8
8 Full lift
Industrial valve
7 L=4.5 mm
Pset=0.3 MPa
7 L=1 mm
6
6
5
5
Ṁ [kg/s]
[kg/s]
4 4
SP water
3 2% air 3
5% air
10% air
2 2
20% air
SP air
1 1
0 0
0 1 2 3 4 5 6 7 8 0 0.2 0.4 0.6 0.8 1
L [mm] β [-]
(a) Mass flow rate versus valve opening for different β. (b) Mass flow rate versus volumetric quality
for different valve openings.
Figure 6.2: Mass flow rate versus β and L in industrial valve at P set =0.3 MPa.
valve and draw some conclusions concerning influence of the valve on two-phase flow
structure. Visualisation
Single-phase
Cavitation-
Cavitation-P=3 barwater flow was investigated in order to visualize cavitation bubbles. The
lowest the valve opening the highest the pressure drop for the same set pressure (largest
the cavitation rate). Therefore, a higher amount of cavitation bubbles has been observed
(Fig. 6.3) for lowest valve opening (L=2 mm).
L=2 mm L=4.5 mm L=7.3 mm
Figure 6.3: Cavitation in SRV for P set =0.3 MPa at valve openings L=2, 4.5 and 7.3 mm.
As next step, two-phase air-water flow has been injected in valve. From first observa-
tions it is remarked that flow visualization images are only possible for very low qualities
(β ≤ 2-3%) since above this percentage, air occupies SRV body and therefore view of the
core is blocked. An example is given for P1 =1.5 bar, L=2.6 mm and β=1.5%. In Fig.
6.4, the upstream and downstream part images are shown. By means of image process-
ing, a rough estimation of the bubble diameter upstream and downstream is obtained and
no significant difference between upstream and downstream bubble diameter is observed
with a mean value of around Db =1.55 mm. Hence, we can conclude that the bubble size
is not considerably altered by the presence of the valve.
In Fig. 6.5, the image taken for the same conditions in the core of the valve is illus-
trated. A remarkable increase in the void fraction is observed for only 1.5% upstream
145
Dbubble≈ 1.7 mm Dbubble≈ 1.4 mm
Figure 6.4: Flow visualization in the safety valve model-upstream (left) and downstream
of the valve (right).
volumetric quality (at injection position). This is due to the sudden pressure drop which
according to Boyle’s law will result in increased volume occupied by the gas. With up-
stream flow regime uniform dispersed bubbly flow, a chaotic flow structure is identified
downstream the valve; counter rotating vortices are created due to the shape of valve
disk (“hat” shape) in the core of the valve and an elicoidal movement of the bubbles in
downstream pipe is observed.
From visualization experiments it is concluded, concerning the flow regime transition,
that for bubbly flow as input, the output of valve is a not known flow pattern. Therefore,
since pressure drop correlations of singularities, presented in previous chapters, have been
established for dispersed uniform bubbly flow both upstream and downstream, they can-
not be used to model the valve as series of singularities (e.g. contraction and sudden
expansion) to predict the pressure drop.
Prediction of possible cavitation position with CFD
In this paragraph a simple CFD computation in axisymmetric 2D case of the industrial

SRV is performed to indicate possible locations in the valve at which cavitation may
occur. More information about the exact settings and conditions of the simulation are
presented in §6.4. In order to identify cavitation, the contours of static pressure near the
valve disk is plotted with values lower than saturation pressure of water at 20 ◦ C i.e. 2340
Pa and is depicted in Fig. 6.6. The upstream pressure P set for all four different valve
lifts tested is equal to 0.3 MPaG. This plot cannot be considered as a precise cavitation
simulation but just an indication on where this phenomenon could appear; hence to obtain
more accurate results in this topic cavitation models provided by CFD codes should be
used.
146
6.3 Optical probe measurements in SRV 147
Figure
3bars-no pst6.5: Flow visualization
correction in the core of the safety valve.
- With pst correction no cavitation
L=1.5 mm L=3 mm
L=4.5 mm
L=7.2 mm
Figure 6.6: Possible location of cavitation appearance predicted by CFD for various lifts
at P set =0.3 MPa
6.3 Optical probe measurements in SRV

The modified facility (LUCY III), presented in §3.2.2, has been used to perform tests with
optical probe in SRV transparent valve. Upstream and downstream profiles are obtained
in two directions as indicated in Fig. 6.7(a) and 6.7(b) respectively. Upstream the valve
co-current vertical flow takes place while downstream there is horizontal flow. Since SRV
modeling requires knowledge of stagnation conditions, tests are focused in upstream void
147
Premiers tests avec sonde optiqu
fraction profiles as well as velocity measurements (results downstream the valve are of
minor importance). In downstream section, half profiles are acquired (from the wall up to
middle of the pipe) due to limitation of length of transversal mechanism; tests are mainly
performed to conclude on how presence of valve influences the bubble size and as a result
the pressure drop.
Zoom
Horizontal 1 Wall
Vertical
Premiers tests avec
1 sonde
0 optique
0 Sonde-VKI
(a) Direction of upstream optical probe SRV profiles (horizontal-vertical).

Vertical-h
Zoom First position
1 Wall
Vertical
0 1
Vertical-half profile
(b) Direction of downstream optical probe SRV half profile (vertical).
Figure 6.7: Direction of upstream-downstream SRV optical probe profiles.
The pressure tested upstream is set at 0.15 MPaG for valve opening L=7.3 mm (full
lift). Two volumetric qualities are investigated; 2 and 9%. For β=2%, flow rate of water
is Qwater =4.5 l/s and for β=9%, Qwater =4.2 l/s. Void fraction profiles upstream the valve
are acquired for these conditions in two directions to verify the uniformity of the flow.
Acquisition time of each void fraction point is 60 s and each profile consists of 12 points
(every 1.8 mm).
In Fig. 6.8 and 6.9, upstream horizontal and vertical void fraction profiles obtained for
148
6.3 Optical probe measurements in SRV 149
2% of air respectively are demonstrated. The shape of the profile is relatively uniform for
horizontal direction with two slight peaks at the two edges of the pipe which is possibly
due to jet effect from gas injector. In vertical profile, higher void fraction is detected for
the part of the pipe towards the wall (support of the valve). Since in this direction, the
profile is perpendicular to gas injector, its influence on void fraction profile is relatively
strong.
Upstream 2 %-horizontal
1.2
1.0
0.8
X [-]
0.6
0.4
0.2
0.0
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4
Void fraction [%]
Figure 6.8: Upstream horizontal void fraction profile in SRV for β=2%.
Upstream 2 %-Vertical
1.2
1.0
0.8
X [-]
0.6
0.4
0.2
0.0
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4
Void fraction [%]
Figure 6.9: Upstream vertical void fraction profile in SRV for β=2%.
Bubble velocity is also measured at each point and average value found for these
conditions is Ububble '15.8 m/s with initial liquid velocity (calculated from flow meter)
Uinitial =12.1 m/s. In order to compare an average bubble velocity measured with the probe
and the initial liquid velocity, the dimensionless slip velocity can be defined:
149
Ububble
s= . (6.1)
Uinitial
This parameter is equal to s=1.3 for β=2% in upstream position. A slip ratio >1 is
expected for upward cocurent vertical flow due to the high density difference between air
and water. Finally, Sauter mean diameter is Dsm '0.87 mm. This value is significantly
different from bubble diameter obtained with visualization (D'1.5 mm) although β is
similar (1.5%). This is due to the higher valve opening and as a result higher liquid flow
rate (4.5 l/s compared to 2.72 l/s tested in visualization experiment). Since very high
velocities were reached for these conditions, the cross correlation between primary and
secondary tip of the probe failed, hence no bubble size distribution information can be
provided.
A volumetric quality of β=9% is tested as well. Void fraction in horizontal and vertical
direction is plotted in Fig. 6.10 and Fig. 6.11 accordingly. For horizontal profile the
shape is similar while for vertical a much more pronounced influence of the gas injector
is noticed. Three strong void fraction peaks which correspond to the three metallic bars
of gas injector form a non-uniform profile. Hence, a significant influence of jet effect for
these conditions is concluded. Average void fraction for β=2% is equal to 3.1%.
Upstream 9 %-horizontal
1.2
1.0
0.8
X [-]
0.6
0.4
0.2
0.0
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5
Void fraction [%]
Figure 6.10: Upstream horizontal void fraction profile in SRV for β=9%.
The bubble velocity is slightly lower than for 2% of air (Ububble '15.1 m/s) possibly
due to coalescence of bubbles forming larger heavier bubbles. Initial liquid velocity is
calculated as Uinitial =11.3 m/s which corresponds to comparatively higher than for β=2%
dimensionless slip velocity of s=1.34. Finally, Dsm '1.55 mm which proves the above-
mentioned conclusion for bigger and therefore with lower velocity bubbles.
Downstream half vertical profiles are also acquired for the same two conditions: 2%
and 9% and results are reported in Fig. 6.12 and 6.13 respectively. Both profiles have
similar shape with a peak at the top of the tube and a slightly more symmetrical profile for
β=2%. From flow visualization no visible stratification at downstream piping has been
observed for these flow conditions. Therefore, no considerable change in void fraction
values can be assumed for the bottom part of the pipe.
150
6.4 Flowforce in SRV 151
Upstream 9 %-Vertical
1.2
1.0
0.8
X [-]
0.6
0.4
0.2
0.0
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5
Void fraction [%]
Figure 6.11: Upstream vertical void fraction profile in SRV for β=9%.
Downstream 2 %-Vertical-half profile
1.0
0.8
0.6
Y [-]
0.4
0.2
0.0
0 0.5 1 1.5 2 2.5 3
Void fraction [%]
Figure 6.12: Downstream vertical void fraction profile in SRV for β=2%.
For the lowest volumetric quality (β=2%), average void fraction is α=2.4% which
corresponds to an increase from upstream void fraction of ' 85% due to expansion of air.
Same conclusion can be drawn for higher volumetric quality (9%) with 80% increase and
average void fraction of 16.2%.
6.4 Flowforce in SRV

The discussion is focused on the physics of the flow (flowforce applied on valve disk) dur-
ing valve operation in compressible, incompressible and two-phase flow. Hydrodynamic
forces on SRV are of crucial importance since their knowledge will permit determining the
spring used with appropriate stiffness. Most valves are initially designed for gas and then
151
Downstream 9 %-Vertical-half profile

1.0
0.8
0.6
Y [-]
0.4
0.2
0.0
0 5 10 15 20 25
Void fraction [%]
Figure 6.13: Downstream vertical void fraction profile in SRV for β=9%.
for their use in water or two-phase flow, coefficients are applied. Therefore, investigation
of force on valve disk for different lift positions under compressible and incompressible
flow will reveal differences on behavior of valve and will provide information on the pos-
sibility of same valves being used in all flow conditions with appropriate adjustments.
The design of transparent valve did not allow direct measurement of force, however
pressure can be measured along the radius of valve disk and after integration along these
points flowforce is extracted (Eq. 3.1). More details have been already provided in
§3.2.1.1. For verification purposes, same measurements were acquired on industrial valve
with force sensor placed at the top of SRV stem. The spring is removed and a specially
designed setup allows fixing the disk at several locations every 0.5 mm. The hydrody-
namic force pressing against the valve disk is recorded with a force sensor. The system
used for flowforce measurement in industrial valve is illustrated in Fig. B.4 in Appendix
B.
6.4.1 Axisymmetric theoretical analysis of flowforce in SRV

A theoretical analysis to approximate the flowforce of SRV disk in simplified axisymmet-
ric system of plate over a nozzle is attempted in this paragraph. We assume the valve disk
of a mass m plated over the nozzle. In order to calculate these forces we consider the
simplified sketch of the system nozzle-disk shown in Fig. 6.14.
Applying continuity and momentum equation for incompressible, inviscid, stationary
flow, we obtain:
Continuity
∂u ∂v
+ =0 (6.2)
∂y ∂x
Momentum
-y axis:
152
d 2x dx
= mSRV
6.4 FlowforceFh in + B + k (h + x) 153
dt 2 dt
R2
Pb
Sb Rb
S2
h S' v, Pb
Outlet
y
u1
P1
Inlet x
D1 S1
Applying continuity and momentum equation for incompressible, stationary flow we

obtain: 6.14: Simplified schematic of valve disk with nozzle.
Figure
Continuity:
∂u ∂v
+ =0  0
0

∂y ∂x 
 ∂u ∂u ∂u 
7 ∂P
ρ  + u + v  = −

− Fy − Fg (6.3)
∂t ∂y ∂x  ∂y
1
-x axis:
 0 0 

 ∂v ∂v ∂v

∂P
ρ  + u + v  = −

− Fx (6.4)
∂t
∂y ∂x  ∂x
The first term of the LHS of Eq. 6.3 and Eq. 6.4 is equal to zero due to stationary
flow. Additionally, axial and radial velocity are constant with x and y respectively hence
∂u/∂x=∂v/∂y=0. The term Fg denotes the gravity force which can be neglected for this
case.
Therefore, we can write:
∂P ∂u
Fy = − − ρu
∂y ∂y
(6.5)
∂P ∂v
Fx = − − ρv
∂x ∂x
The flowforce applied from the fluid to the valve disk is equal to Fy since the sucking
forces from the fluid on -x axis are canceling each other. Hence,
πh i
Fy = − P1 S 1 + ρu21 S 1 + Pb · D22 − (D2 − Db )2 (6.6)
4
153
The velocity u can be replaced by:
q1
u= , (6.7)
S1
where q1 the flow rate charged upstream.

Thus, the final theoretical formula for hydrodynamic flowforce applying on valve disk
is given by:
q21
!
Fh = − P1 S 1 + + Pb · πRb (2R2 + Rb ) (6.8)
S1
Comparing experimental results with theoretical calculation for P1 =0.3 MPa, we ex-
tract the plot for single-phase water flow shown in Fig. 6.15. The overall deviation of this
theoretical calculation from experimentally determined flowforces for several set pres-
sures is ' 11% (with minimum and maximum 6 and 15% respectively).
3 bar-15 crans
300
250
200
F [N]
150
100
50
Experimental
Theoretical
0
0 1 2 3 4 5 6 7 8
L [mm]
Figure 6.15: Comparison of experimental and theoretical hydrodynamic force of SRV

versus valve opening for P1 =0.3 MPa in water flow.
If a two-phase flow mixture is considered the mixture density will be given by Eq. 2.18.
Hence, same methodology can be applied replacing the density of liquid ρL with ρ∗ .
The case of upstream P1 =0.3 MPa and two-phase mixture of β=20% is considered and
comparison between model and experiments for these conditions is presented in Fig. 6.16.
A satisfactory agreement withing engineering purposes is also observed for two-phase
flow with average deviation of 14.95% for all valve openings.
From this analysis, we can conclude that flowforces on valve disk can be determined
considering a relatively simple theoretical model with an acceptable accuracy for most
tested conditions both for single-phase water flow and two-phase mixtures.
154
3 bar-15 crans-20 % air

250
200
150
F [N]
100
50
Experimental
Theoretical
0
0 1 2 3 4 5 6 7 8
L [mm]
Figure 6.16: Comparison of experimental and theoretical hydrodynamic force of SRV

versus valve opening P1 =0.3 MPa in two-phase flow conditions (β=20%).
6.4.2 Experimental investigation

In Fig. 6.17, force applied on the valve disk versus valve opening for set pressure P1 =0.2
MPa in the transparent and industrial valve is shown. For the case of transparent valve,
the set conditions, for which comparison with industrial valve is attempted, correspond to
pressure difference between upstream and back pressure (P set -Pb =0.2 MPa). The position
of adjustment ring is fixed at 15 notches (zero vertical distance between valve seat and
adjustment ring) for all cases studied during this project. Detailed explanation of the
relative position of valve seat and adjustment ring is given in §3.2.1.3. The exact position
of 15 notches is shown in Fig. 3.19(a).
A relatively good agreement between the two cases is noticed; deviations between 15
and 30% can be caused to errors coming from calculation of integration of pressure of only
three points along the disk radius for transparent valve. Additionally, the influence of back
pressure and longer nozzle for the case of transparent valve are possible explanations of
this discrepancy. Variation of flowforce is noticed for the lower valve lifts for both cases
due to the position of adjustment ring. The section between valve seat and ring create
a small annular cavity hence recirculation zone is formed at this position affecting the
opening characteristics of the valve.
Measurements with varying adjustment ring positions have been performed at the
facilities of CETIM to reveal the influence of this parameter on flowforce curve (Corbin
and Chabane (2009)). The plot shown in Fig. 6.18 demonstrates flowforce versus valve
opening at four different positions of adjustment ring and for P set =1.21 MPa. The lower
the number of notches, the higher the position of the ring thus the greater its influence on
flowforce (bigger cavity). This phenomenon is more pronounced for the lowest pressures
since lower fluid velocities occur and the fluid jet impacts more severely at the system
valve seat-ring.
Flowforce characteristic curves (set pressure P, valve lift L, flowforce F) of SRV for
155
Comparaison soupape transparente-metallique-eau 15 crans
200
Pset=0.2 MPa
180
160
140
120
F [N]
100
80
60
40
Transparent valve
20
Industrial valve
0
0 1 2 3 4 5 6 7 8
L [mm]
Figure 6.17: Force applied on the valve disk versus valve opening for transparent and
industrial valve. Efforts sur clapet soupape 1''1/2 G 3"
1200
1000
800
F [N]
600
400
F (5 notches)
F (10 notches)
200
F (15 notches)
F (20 notches)
0
0 1 2 3 4 5 6 7 8
L [mm]
Figure 6.18: Influence of adjustment ring position on flowforce for P set =1.21 MPa.
air and water flow are plotted in Fig. 6.19. Force variation with valve lift at several set
pressures is demonstrated. For the case of incompressible flow (Fig. 6.19(a)), increase of
flowforce versus valve lift is clearly more abrupt at the smallest openings while for largest
openings it follows an asymptotic trend. On the other hand, force is linearly increasing
with pressure for a given lift.
Flowforce characteristic plot under compressed air flow conditions is depicted in
Fig. 6.19(b). In this case, a completely different performance of flowforce compared
to water is detected. More uniform distribution of F=f(P, L) is found with a gradual force
increase with both pressure and valve lift. Pop-action of the valve appears in gaseous con-
ditions, since a relatively high fluid force acts on the disk even for the smallest openings;
this means that the valve opens suddenly (low lift) and then the force increase is more
156
smooth. These conclusions are in accordance with results presented from Föllmer and
Schnettler (2003).
Force [N]
(a) Flowforce characteristic for incompressible flow.
Force [N]
(b) Flowforce characteristic for compressible flow.
Figure 6.19: 3D Flowforce characteristic (P, L, F) plots in industrial SRV.
Uniform two-phase bubbly flow is injected upstream the valve in order to compare
the force applied at the disk for air-water mixture with incompressible and compressible
behavior. For the case of transparent valve, the force applied on the valve disk is calculated
157
by integrating the pressure measured at the three pressure taps located along the radius
of the disk. Hence, for P set =0.27 MPa, the force is plotted against the opening of valve
for single (water) and two-phase flow of varying volumetric quality β (Fig. 6.20(a)). An
inversion of force value (from higher to lower) between air and water flow is remarked
at L'3 mm. In order to investigate this behavior, measurements have been repeated for
same flow conditions in industrial SRV with force sensor.
In Fig 6.20(b), flowforce measured in industrial valve is plotted against valve opening
for pure air, pure water and 2, 5, 10 and 20% air-water flow mixtures at set pressure
P set =0.3 MPa. A minimum of the force is noticed for water at 1.7 mm valve lift. This is
due to the presence of adjustment ring, which at this position creates small cavity resulting
in fluid entrainment. Hence, the section of the disk that pressure is applied, and as a
consequence the flowforces at this position are diminished.
Inverse flow behavior between air and water is noticed at 3 mm valve lift. For the
lowest openings force for air flow is superior than for water while for highest openings
vice versa. Numerical simulations are performed for same valve and conditions with the
purpose of explaining this behavior and results are presented in the following paragraphs.
Two-phase flow tends toward compressible behavior for the highest lifts with equal forces
of air and 20% air mixture. Below the flowforce inversion position, although two-phase
mixtures are slightly biased toward air forces, they still remain closest to incompressible
behavior.
Figure 6.21(a) illustrates the influence of the gaseous phase on the force applied at the
disk. Up to volumetric quality of 10%, no significant difference is noted. Hence, in order
to identify the small difference of the force between single and two-phase flow, the ratio
∆F is defined:
FT P − FS P,water
∆F = · 100% (6.9)
FS P,water
Distinctive deviations are demonstrated in Fig. 6.21(b) for industrial SRV. A maxi-
mum discrepancy ∆F of 60% is noted for 1.5 mm valve opening (influence of adjustment
ring), then becomes zero for 3 mm and afterwards becomes negative down to 20% from
5.5 mm to the maximum lift (7.2 mm). For the highest air qualities (β=20%) flowforce
values are nearly matching compressible flow behavior.
In order to identify the reason for this inverse behavior between compressible and
incompressible flow, higher pressures are tested (0.6 and 1.1 MPa). The force plots for all
tested conditions are presented in Fig. 6.22. It is remarkable that the position of inverse
behavior is changing with pressure. The higher the set pressure, the greater the valve
opening for which force is inversed. The pressure ratio for the valve is expressed as
n=Pb /P set and since for the examined case fluid is always discharged at the environment,
Pb is always equal to unity (Patm ).
As it has been explained in §2.4.6, in SRV and nozzles rather often the flow is “blocked”
meaning that is not increasing with further decrease of downstream pressure. The flow
rate and pressure for which the latter situation occurs are called critical flow rate critical
pressure respectively.
We recall the definition of critical pressure ratio (ncrit =Pcrit /P set ) for air:
158
Soupape transparente 15 crans
350
Transparent valve
Pset=0.27 MPa
300
250
F [N] 200
150
Water
100
2 % air
5 % air
50 10 % air
Air
0
0 1 2 3 4 5 6 7 8
L [mm]
(a) Flowforce applied 3inbar-soupape SRV at P set =0.27 MPa.
transparentmetallique
300
Industrial valve
Pset=0.3 MPa
250
200
F [N]
150
100 SP water
2% air
5% air
50 10% air
20% air
SP air
0
0 1 2 3 4 5 6 7 8
L [mm]
(b) Flowforce applied in industrial SRV at P set =0.3 MPa.
Figure 6.20: Force applied on the valve disk versus valve opening for water, air and
various air-water mixtures
!(k/(k−1))
2
ncrit = (6.10)
k+1
where k the ratio of specific heats.
The difference of the actual pressure ratio from pressure ratio at which chocked flow
occurs (critical pressure ratio) can be illustrated by defining the following ratio:
ncrit − n
∆ncrit =
· 100% (6.11)
ncrit
In Fig. 6.22, the values of ∆ncrit are indicated for each of the three sets of curves
corresponding at three different upstream pressures. From the graph, it is revealed that
159
3 bar-soupape transparente-subcritique
25
2% air
5% air
20 10% air
14% air
15 20% air
10
∆F [%]
5
-5
-10 Transparent valve

Pset=0.3 MPa
-15
0 1 2 3 4 5 6 7 8
L [mm]
(a) Deviation of flowforce in transparent SRV.
70
Industrial valve 2%
60 Pset=0.3 MPa 5%
10%
50 20%
Air
40
F [%]
30
20
10
-10
-20
0 1 2 3 4 5 6 7 8
L [mm]
(b) Deviation of flowforce in industrial SRV.
Figure 6.21: Relative discrepancy between flowforce of air and two-phase and single-
phase water at P set =0.3 MPa.
enhanced position of inverse force corresponds to increasing ∆ncrit ratio. Therefore, it can
be concluded that alteration of force values between air and water is directly linked to
change of chocked flow position; for high ∆ncrit ratio (n << ncrit ) inverse flowforce occurs
at high lifts and vice versa.
6.4.3 CFD simulations

Numerical simulations are carried out with commercial code Fluent for air and water flow
in order to validate experimental results. A simplified geometry is chosen, considering
that the flow close to the valve outlet is not significantly affected by flow impinging on
valve body and therefore the 2D axisymmetric grid shown in Fig. 6.23 is used. Exper-
160
1200
2% 5%
7% 10% ∆ncrit=85 %
1000 20% Air
Water
800
1.1 MPa
F [N] ∆ncrit=75 %
600
400
0.6 MPa ∆ncrit=57 %
200
0.3 MPa
0
0 20 40 60 80 100
Valve opening [%]
Figure 6.22: Comparison of inverse flowforce position for different set pressures in in-
dustrial SRV.
imental flow visualization in transparent setup for water flow with low amount of small
bubbles injected, used as tracers, indicated that axisymmetric flow could be assumed.
The lateral hypothesis agrees with the remarks of Vallet et al. (2010) for the same type of
valve.
The results from computations are focused on the disk part since flowforce informa-
tion is required, hence the mesh is adapted close to the valve disk. The mesh size is '
85000 cells, the flow is turbulent for all cases and viscous model used is the two equation
standard k-ω with pressure based coupled solver. For comparison reasons, one position of
the adjustment ring (shown in the zoomed grid in Fig 6.23) was chosen for all simulations
which is the same used for the whole experimental campaign.
In Table 6.1, experimental and numerical flow conditions are summarized.
Table 6.1: Flow conditions for SRV tests.
P set [MPa] L [mm] β [%]

Exp/Air 0.07-1.1 0.5-7.2 (full lift) 100
Exp/Water 0.07-1.1 0.5-7.2 0
Exp/Air-water 0.3 / 0.6 1.2-7.2 2-20
CFD/Water 0.2 / 0.3 / 0.6 / 1.1 1.5 / 3 / 4.5 / 7.2 0
CFD/Air 0.2 / 0.3 / 0.6 / 1.1 1.5 / 3 / 4.5 / 7.2 100
6.4.3.1 CFD results

More insight on the preliminary previously mentioned phenomena is given with analy-
sis of CFD results. Computations are carried out in original vale axisymmetric mesh for
compressible and incompressible fluid and same flow conditions. In reality, when valve
161
Figure 6.23: Axisymmetric grid of industrial SRV for CFD computations.
operates in water, cavitation (formation of vapor bubbles) occurs downstream the valve
due to local depressurization (liquid pressure drops below saturation pressure). In the sim-
ulations this phenomenon is not taken into account to decrease complexity of the problem
and computational time.
6.4.3.1.1 Water flow The contours of total pressure close to the valve disk at four dif-
ferent valve openings (7.2, 4.5, 3 and 1.5 mm) and streamlines of water flow for P set =0.6
MPa are shown in Fig. 6.24. Flow contracts upstream the valve, resulting in flow acceler-
ation, then passes through the very narrow section between the disk and valve seat at the
highest velocity region and then, due to change of flow direction, recirculation regions are
created with fluid impacting on valve body. These regions occupy large areas interacting
with system valve disk-seat for highest lifts (7.2 and 4 mm) thus generating discharged
fluid jet of ' 45 ◦ angle. On the other hand, for lowest lifts (3 and 1.5 mm), smallest fluid
recirculation region is entrained in the area defined by adjustment ring and valve disk.
Therefore, fluid is reflected with a 90 ◦ angle approximately.
Total pressure evolution is also illustrated in Fig. 6.24. Lowest valve lifts produce
high pressure concentrated in the middle of the disk while for larger openings, pressure
is more uniformly distributed along the valve disk (higher pressure gradient). Hence, as
it was demonstrated in Fig. 6.19(a), at lowest lifts, sudden increase of force with rapid
opening of the valve occurs, while for highest lifts, a more gradual increase of force takes
place (asymptotic shape of flowforce curve). Low pressure areas define possible position
of cavitation occurrence. These regions are noticed very close to the disk for L=1.5 mm
whilst for full lift (L=7.2 mm), they are located downstream the valve disk. The latter
observations are consistent with conclusions of Vallet et al. (2010).
Influence of cavitation can be demonstrated, as it was shown in Vallet et al. (2010),
by plotting the following ratios; F/∆P and F/Q2 versus valve opening (Fig. 6.25). Force
162
Figure 6.24: Total pressure contours for water flow at P set =0.6 MPa.
to pressure drop ratio calculated with CFD does not flawlessly fit with experimental re-
sults for lowest valve opening whereas for highest lift an excellent agreement is remarked.
Same statement can be made for the plot of ratio F/Q2 . The shape of both curves confirm
the sharp increase of discharged water flow section for the lowest lifts and a more grad-
ual further increase of the section at more enhanced valve openings. Slight discrepancy
between measured and computed values for the lowest lift (1.5 mm) can be explained by
experimental uncertainty on the exact adjustment ring position which will be discussed
more in detailed in the next paragraphs. At this point, it worths mentioning that when high
cavitation rates take place in SRV, the latter should be taken into account for the correct
sizing of the valve.
6.4.3.1.2 Air flow In compressible flow, contours of density close to the valve disk and
flow streamlines are shown in Fig. 6.26 for P set =0.6 MPa at various valve lifts. In the four
graphs, solid thick white line denotes sonic iso-contours (Mach number Ma=1). Contrary
163
Water-P=3 bars-cavitation influence Water-P=3 bars-cavitation influence
1000 1.8
CFD
900 1.6 EXP
800 1.4
700 1.2
F/Q [Nh /m ]
6
F/∆P [mm ]
2
600
1.0
2
500
0.8
2
400
0.6
300
200 0.4
CFD 0.2
100
EXP
0 0.0
0 1 2 3 4 5 6 7 8 0 1 2 3 4 5 6 7 8
L [mm] L [mm]
(a) Experimental-CFD values of F/∆P. (b) Experimental-CFD values of F/Q2 .
Figure 6.25: Influence of cavitation on flowforce versus valve opening
to water flow, the discharged fluid jet seems to be only slightly influenced by valve lift
with an approximate 45 ◦ angle for all cases. Moreover, recirculation regions are mainly
located downstream the system valve disk-seat-adjustment ring and at approximately the
same position for the four valve openings which explains the similar shape of reflected
jet.
Chocked flow takes place at different positions for each valve lift as it is
indicated in the graph presented in Fig. 6.26.
For full lift (L=7.2 mm), the position of sonic line is located exactly at the edge of
valve seat which is the minimal annular section for this case. For lower opening (L=4.5
mm), minimum section is moved further downstream the valve and for L=3 and 1.5 mm,
sonic position is located at the annular huddling chamber created between valve disk and
adjustment ring. The section of the disk that force applies is influenced by sonic line
position. Therefore, we can conclude that the section of effective disk force is modi-
fied according to the displacement of the sonic line. The previously analyzed different
flow behaviors of air and water provide an explanation of inverse flowforce noticed in
Fig. 6.20(b) and Fig. 6.22.
6.4.3.2 Comparison flowforce experiments-CFD

Experimentally measured force acting on valve disk for air and water is compared with
CFD results for P set =0.3 MPa (Fig 6.27). For the lateral case, force is calculated from
the CFD code as the sum of viscous and pressure forces. In this selected condition, to
achieve a better accuracy, a finer grid of 1.45 million cells is used; a dimensionless wall
distance y+ ' 20 is obtained in order to attempt resolving the viscous sublayer of the
flow. For compressible flow, experimental results and CFD calculations agree within
8%. For pressures 0.2, 0.6 and 1.1 MPa, simulations are also in good accordance with
measurements with maximum deviation of 22% observed at 1.1 MPa and L=4.5 mm.
When considering incompressible flow, experimental results agree within 3% with
CFD for most lifts with the exception of lowest opening L=1.5 mm for which slightly
lower force is predicted. One possible reason for this minor disagreement is experimental
164
6bars air
6bars air
+stream
+stream
Figure 6.26: Density contours for air flow at P set =0.6 MPa. Solid white thick line indi-
cated sonic position.
uncertainty on the position of adjustment ring, which only influences water flow as ex-
plained in the previous sections, and could significantly affect the flowforce. A parametric
study of adjustment ring position has proved that a trivial difference of 0.23 mm on ring
position can result in 50% deviation of flowforce. The uncertainty on the location of ring
is estimated for this case of the order of 0.11 mm, corresponding to 52% of the actual
ring displacement. Hence, a considerable ambiguity on measurement of adjustment ring
position is highlighted.
To summarize, we can conclude that the previously presented preliminary CFD cal-
culations are adequately accurate to predict flowforce both for compressible and incom-
pressible flow. For the latter conditions, at the smallest lift tested, an alteration of the disk
force is remarked due to the lack of experimental confidence on the exact ring location.
165
Std k-omega, y+ ~ 20, Grid 1.45M cells
300
250
200
F [N]
150
100
Water-CFD
50 Water-Experimental
Air-CFD
Air-Experimental
0
0 1 2 3 4 5 6 7 8
L [mm]
Figure 6.27: Experimental-CFD flowforce for air and water at P set =0.3 MPa.
6.4.3.3 Influence of adjustment ring position
The valve opening characteristics can be modified through the use of adjustment ring.
More details on its role have been given in §2.4.2. In the present paragraph an attempt
to quantify, by performing CFD computations, the influence of the position of adjustment
ring on flowforce is presented. Hence, the area of the disk at which high static pressure
changes occur is located. The latter, called hereunder the “sensible area”, is the critical
section for the determination of the flowforce.
Numerical simulations have been initially performed for 15 notches (valve seat and
ring position at the same height) with P set =0.3 MPa at L=1.5 mm. An additional two ring
positions; 0 and 20 notches were tested for comparison with reference value of 15 notches.
More explanation of notches definition and its unit length are given in §3.2.1.3. For the
radial distance defined by the disk from 0 to approximately 14 mm no static pressure
changes are indicated while from 14 mm up to the disk radius (sensible area) large static
pressure variations have been remarked.
A theoretical calculation of the flowforce has been carried out with variation of the
static pressure only in the sensible area by ± 50 and 90%. The results are demonstrated
in Fig. 6.28(a). The calculated flowforce for each case is indicated with side arrows.
Variations of 12% and 23% for 50 and 90% static pressure changes are concluded. Addi-
tionally, in order to quantify the force variation for different ring positions, static pressure
versus disk radius is plotted for 3 settings; 0, 15 and 20 notches shown in Fig 6.28(b).
Compared to reference ring location of 15 notches, force varies 52% and 73% for 20 and
0 notches respectively. It is remarkable that by moving the adjustment ring only for few
millimetres (as indicated in the figure), significant changes in flowforce occur. Given the
lack of precision in the ring regulation, an important experimental uncertainty is intro-
duced in the measurement of the exact ring position.
166
6.5 Pressure drop in SRV 167
Therefore, it can be concluded that valve designers and users should be take special
care at the adjustment ring, used for the crucial smallest openings of the valve.
L=1.5 mm
Ssensible
Ssensible
(a) Flowforce theoretical calculation for varying P st at the sensible area.
F=73 %
L=1.5 mm
F=52 %
• 15 notches: + 0.0635 mm
• 0 notches: + 0.296 mm
• 20 notches: - 0.588 mm
(b) Flowforce CFD computations for various adjustment ring locations.
Figure 6.28: Influence of adjustment ring location on disk flowforce.
6.5 Pressure drop in SRV

Pressure drop coefficient of the valve is extracted for different test conditions. Static
pressure is measured at different locations of the industrial SRV. Total pressure at each
point is then calculated as the sum of static pressure measured and mean dynamic pressure
computed at different sections along the valve. The final pressure drop is defined as
167
the difference of total pressure from reference pressure tap “nozzle 1” upstream and the
pressure tap from the valve “downstream far” (as shown in Fig. 3.14(a) in §3.2.1.1).
A set pressure of 0.3 MPaG and 0.15 MPaG is fixed for different valve openings and
pressure drop is determined. An equivalent pressure drop coefficient ζ is established. The
definition of this coefficient is based on that used for singularities (Chapters 4 and 5).
Ptot tot
set − Pb
ζ= (6.12)
1
ρ J2
2 L L
Pressure drop coefficient versus volumetric quality (up to 20%) for P set =0.15 MPaG
is plotted in Fig. 6.29. The total mass flow for each point is indicated as well. The lowest
the mass flow rate the highest the pressure drop coefficient. Furthermore, lower valve lift
results in lower mass flow rate passing from the valve but highest pressure drop since the
flow obstruction is higher. Volumetric quality of air β=20% produces an increase in ∆P
of ' 25% to 30% compared to water flow for full lift. Therefore, the need for a different
design (sizing) of the valve in two-phase flow is concluded.
In Fig. 6.30 pressure drop coefficient function of volumetric quality of air for three
different valve openings; L=5.5, 6.5, 7.3 mm are compared for two set pressures; 0.15 and
0.3 MPaG. Coefficient ζ is found to be almost independent on upstream set pressure of
the valve. Factors influencing the pressure drop is flow obstruction (valve lift), air quality
and upstream liquid Re number as will be discussed in §6.7.
The effect of air content on pressure drop is shown in Fig. 6.31. The dimension-
less pressure drop ΦL is plotted versus air quality for two pressures 0.15 and 0.3 MPa at
full lift. The results for both pressures lie upon the same fitting curve shown in the fig-
ure. Increase in pressure drop coefficient reaches 100% for the maximum quality tested
(β '50%), hence a ΦL =2 as it is indicated in Fig. 6.31. Results are compared with corre-
lation of Chisholm (1971) presented in §2.4.1, Eq. 2.39. The coefficient B=1.5 for safety
valves is used. Very good prediction of the two-phase multiplier is remarked especially
for the lowest qualities while for higher two-phase mixtures slight deviation is noticed
(possibly due to uncertainty of the measurements performed with nitrogen).
Three dimensional plot of pressure drop coefficient-volumetric quality and valve lift is
shown in Fig. 6.32 for upstream pressure 0.15 MPa. An abrupt raise of ζ with increasing
volumetric quality is remarked for the lowest valve openings (L=3.5 mm) even at low
qualities. On the other hand, for high valve lifts a quasi-linear increase of ∆P coefficient
with β is noticed. Therefore, influence of air on pressure drop and as a consequence on
discharged mass flux is more pronounced for the lower lifts. The latter reveals the opening
characteristics of SRV in two-phase mixture which even for low qualities approach more
to compressible flow behavior (pop-opening of the valve).
6.6 Mass flux through SRV-discharge coefficient

The correct sizing of SRV is of crucial importance for industrial safety. Undersized valves
can cause insufficient fluid discharge and therefore excessive pressure in the equipment
under protection. Hence, there will be a strong possibility of explosion. On the other
hand, oversized SRV will cause overflow, compared to expected discharged flow rate, in
the undersized downstream piping system and thus disastrous consequences may result.
168
6.6 Mass flux through SRV-discharge coefficient 169
3.5
Pset=0.15 MPa
3.0
Ṁ=2.92 kg/s
∆P coefficient ζ [-]
Ṁ=3.06 kg/s
2.5 Ṁ=3.16 kg/s
Ṁ=3.33 kg/s
Ṁ=3.48 kg/s
Ṁ=3.50 kg/s Ṁ=3.18 kg/s
Ṁ=3.35 kg/s
2.0 Ṁ=3.48 kg/s
Ṁ=3.67 kg/s
Ṁ=3.81 kg/s
Ṁ=3.88 kg/s
1.5
L=7.3 mm
L=5.5 mm
1.0
0 5 10 15 20
Figure 6.29: Pressure drop coefficient ζ at full lift and L=5.5 mm versus
volumetric quality for P set =0.15 MPaG.
3.5
3.0
∆P coefficient ζ [-]
2.5
2.0
Pset=0.3 MPa-L=7.3 mm
1.5
1.0
0 5 10 15 20
Figure 6.30: Verification of independence of pressure drop coefficient

on set pressure.
The aforementioned scenarios point out the necessity for deep analysis of SRV use under
several conditions. A detailed description of different methodologies adopted to predict
the mass flux through SRV have been presented in §2.4.7.
In order to calculate the flow discharge through SRV, ω parameter should be estimated.
The first term of ω parameter (Eq. 2.75) corresponds to upstream two-phase mixture while
second term denotes the presence of cavitation bubbles or complete flashing flow in high
depressurization rates downstream the valve.
Since the criterion to discriminate whether cavitation occurs or not, is only visible
and therefore not very accurate, an estimation of flow discharge with different prediction
169
3.0
y = 3.1258x2 + 0.5593x + 1.039
R2 = 0.9927
2.5
2.0
ΦL [-]
1.5
1.0
P=1.5 bar
0.5 P=3 bar
Chisholm (1971) correlation L=7.3 mm

0.0
0.0 0.1 0.2 0.3 0.4 0.5 0.6
Volumetric quality β [-]
Figure 6.31: Dimensionless pressure drop ΦL function of volumetric quality of air at full
lift for P set =0.15 and 0.3 MPaG. Comparison with Chisholm (1971) correlation.
ζ [-]
Figure 6.32: 3D plot of β-L-ζ of transparent SRV at P set =0.15 MPaG.
170
methods is distinguished in two cases; flashing and non-flashing use of ω parameter.
6.6.1 Non-flashing-“Frozen” two-phase flow

The mass flux through the transparent valve for different inlet stagnation qualities at max-
imum lift is measured experimentally and then calculated with ω-methodology for non-
flashing conditions; it can also be referred as “frozen” flow. In Fig. 6.33 the result is
depicted for two set pressures; 0.15 and 0.3 MPa. For both pressures subcritical regime
condition is applied.
A slight decrease of mass flux with increasing quality is monitored for the lowest
stagnation qualities (0.01%) reaching asymtotically the mass flow of pure gas flow at
x=100%. Same trend of predicted and experimental curves is observed for both pressures
with overprediction of mass flux from ω-methodology. This appears to be logical since
a certain amount of vapor bubbles is produced due to cavitation in reality (not taken into
account for “frozen” flow) and therefore the discharged mass flux is lower for same inlet
stagnation air quality. L=7.3 mm
Kdl=0.73 , Kdg=0.96
25000
P=0.3 MPa
Omega method-P=0.3 MPa
P=0.15 MPa
Critical mass flux G [kg/m s]
Omega method-P=0.15 MPa

20000
2
15000
10000
5000
0
0.0000 0.0001 0.0010 0.0100 0.1000 1.0000
Stagnation quality x [-]
Figure 6.33: Critical mass flux measured and calculated with ω-method for non-flashing
conditions in transparent SRV.
The ω-methodology is also applied for industrial valve tests at set pressures; 0.3, 0.6,
0.7 and 0.9 MPa under critical flow conditions (Fig. 6.34). Comparison of experimen-
tal to calculated results leads to similar conclusions with transparent case. A significant
overprediction of mass flux from ω-method for the lowest stagnation qualities and bet-
ter matching of experimental results with theoretical calculations for highest qualities is
detected.
171
Additionally, prediction of mass flux with API RP 520 (2000) methodology is at-
tempted (presented in §2.4.7.5). This model slightly underestimates the flow rate for the
highest qualities with predictive accuracy being improved increasing set pressure; for 0.9
MPa a very good agreement with experimental findings is remarked. For low qualities of
air (close to pure liquid) the method predicts perfectly the mass flux.
At this point, the definition of discharge coefficient Kd is recalled:
ṁexp
Kd = (6.13)
ṁnozzle
where ṁexp the experimentally determined mass flow rate through SRV and ṁnozzle the
mass flow rate through isentropic ideal nozzle.
Therefore, the ratio of experimental to calculated from ω-method mass flow rate cor-
responds to discharge coefficient and is indicated in the plot for each pressure. For lowest
qualities, the values of Kd tend towards liquid coefficient (Kdl =0.734) while for the high-
est amount of air, Kd approaches the discharge coefficient for pure gas (Kdg =0.960).
4
x 10 P=0.3 MPa−metallic valve 4
x 10 P=0.6 MPa−metallic valve
2.6 3.6
Experimental
2.4 HEM 3.4
Mass flux [kg/m2s]
Mass flux [kg/m s]
API 520 Experimental

2
3.2 HEM
2.2
3 API 520
0.738 0.737
2
2.8
1.8
2.6 0.85
1.6 2.4
0.892
1.4 −5 −4 −3 −2
2.2 −6 −5 −4 −3
10 10 10 10 10 10 10 10
Mass quality [−] Mass quality [−]
4
x 10 P=0.7 MPa−metallic valve 4
x 10 P=0.9 MPa−metallic valve
3.8 4.4
3.6 4.2
Experimental
Mass flux [kg/m2s]
Mass flux [kg/m s]
4
3.4
2
HEM Experimental
API 520 3.8
3.2 HEM 0.752
0.735
3.6 API 520
3
3.4
0.819 0.831
2.8
3.2
2.6 3
2.4 −6 −5 −4 −3
2.8 −6 −5 −4 −3
10 10 10 10 10 10 10 10
Figure 6.34: Measured and calculated mass flux with ω-method for non-flashing condi-
tions under various set pressures in metallic valve.
Starting from definition of Kd (Eq. 6.13) and expressing the theoretical mass flow rate
with continuity equation and fluid velocity function of ∆P and ρ, we obtain following
equation:
172
ṁexp
Kd = q (6.14)
ρA 2∆P ρ
And with pressure drop coefficient defined as: ζ=∆P/0.5ρU2 , the coefficient Kd can
be written in the form:
ṁexp

Kd =
√ζ
(6.15)
ρAU

Therefore, the discharge coefficient can be calculated through pressure drop coeffi-
cient with the following formula:
1
Kd = √ (6.16)
ζ
It is important to point out that pressure drop coefficient for the valve in this case is
not defined as in Eq. 6.12; superficial velocity of the two-phase mixture should be used
for dynamic pressure computation:
Ptot tot
set − Pb
ζ= (6.17)
1
2
ρJtot
2
When valve discharges in atmosphere, Eq. 6.16 gives Kd =f(ζ). In contrast, when
SRV is connected with downstream piping, back pressure Pb is present and an additional
regular pressure drop coefficient ζ1 should be added in Kd calculation:
1
Kd1 = √ (6.18)
ζ + ζ1
where regular pressure drop coefficient is given by:
l
ζ1 = λ (6.19)
D
with λ the Darcy friction factor, l and D the length and diameter of downstream piping
respectively.
The two-phase flow discharge coefficient is calculated as ratio of measured discharged
flux over G estimated with ω-methodology and ζ calculation for each two-phase flow
condition (Eq. 6.20). In Fig. 6.35 Kd is plotted against volumetric quality for two set
pressures in transparent valve; 0.15 and 0.3 MPa and five set pressures in industrial valve;
0.3, 0.5, 0.6, 0.7, 0.9 MPa.
Gexp
Kd = (6.20)
Gcalc (ω method or ζ calculation)
The single-phase experimentally determined discharge coefficients for water and air
(subcritical and critical condition) are indicated. For transparent valve, a linear increase
of Kd with air content is noticed, from pure water flow (water discharge coefficient Kdl )
up to β=100% (air discharge coefficient Kdg ). On the other hand, for industrial valve
the curve asymptotically reaches Kdg . Comparison with correlation from Lenzing et al.
(1998) proves a relatively good agreement with transparent valve results.
173
An overestimation of discharge coefficient is remarked for calculation with ω-method

both for transparent (Fig. 6.35(a)) and industrial valve (Fig. 6.35(b)). Therefore, a lower
discharge mass flux is predicted since Kd is the ratio of measured over predicted mass flux.
This is possibly due to partial cavitation occurring (higher amount of bubbles) which is
not taken into account in this case. For industrial valve, discrepancy is more significant
compared to transparent SRV since higher pressures are tested (the highest the pressure,
the largest the cavitation rate).
Additionally, the lower values of Kd for transparent valve compared to industrial valve
are explained by the presence of back pressure in first case which creates a higher pressure
drop coefficient (Eq. 6.18) and therefore lower discharge coefficient.
Finally, the discharge coefficient for transparent valve is calculated with ω-method at
several valve openings for air, water and three different two-phase flow mixtures; β=2, 5
and 10% air. The plot for P set =0.3 MPa is shown in Fig. 6.36. Higher discharge coefficient
is noticed at the largest valve openings for water compared to air while for lowest lifts this
tendency is inversed; transition point is L=3.5 mm. Since Kd is inversely proportional to
pressure drop, this means a higher pressure drop in high lifts for water and lower pressure
drop for smaller lifts and vice versa for air. Flowforce plots for same conditions resemble
with previously analyzed behavior with inverse signs of air and water Kd .
Two-phase mixtures follow mainly water trend up to 4 mm of valve opening while
for highest lifts they start approaching Kd of gas. At full lift and only 10% of air dis-
charge coefficient is increased by 9.5% compared to water and is 12% lower than gas
Kd . In §6.4.3.1 an explanation of this phenomenon is given; higher flow section for air at
low valve lifts results in higher pressure drop and smaller section at high valve openings
creates a lower pressure drop compared to water.
From experimental observations presence of few cavitation bubbles downstream the
valve was observed for P set ≥0.15 MPa. Hence, the use of flashing flow formulas to
estimate ω parameter appears to be more appropriate.
6.6.2 Flashing two-phase flow

In reality partial cavitation due to depressurization occurs downstream the valve. In this
occurrence, ω parameter with flashing flow is calculated according to Eq. 2.75. Further-
more, mass flux is estimated with Homogeneous Non-Equilibrium Model of Diener and
Schmidt using the methodology that has been demonstrated in §2.4.7.3.
Calculation of compressibility factor ωDS requires knowledge of sensitive parameter
α (boiling delay exponent). For long nozzles and safety valves recommended value is
α=2/5 while for short nozzles and control valves appropriate value is 3/5. In the present
investigation transparent and industrial safety valve of type 1 1/2" G 3" are studied. For
transparent model a long nozzle is used to discharge the fluid upstream the valve while
for metallic valve, the nozzle can be characterized as short (Fig. 3.11 in §3.2.1.1). Ap-
plication of the model for both values in each case (transparent and industrial SRV) has
confirmed the aforementioned conclusion.
For transparent valve upstream pressures of 0.15 and 0.3 MPa are tested and results
are shown in Fig. 6.37 and Fig. 6.38 respectively. Experimental and calculated discharged
mass flux versus volumetric quality β are plotted in the figures.
An important underestimation of ' 50% is noticed for ω-methodology. Therefore,
174
Kd2ph-Omega method-zeta calculation
1.10
Kdg,cr=0.960
1.00
0.90
Kdg,subcr=0.914
KdTP [-]
0.80
0.70
Kdl=0.734
0.60
Transparent SRV-ω method
0.50 Transparent SRV-ζ calculation
Lenzing et al. [1998]-subcritical
0.40
0 10 20 30 40 50 60
Kd2ph-Omega method-zeta calculation
(a) Kd calculated with ω-method and ζ calculation in transparent SRV/subcritical regime.
1.10
Kdg,cr=0.960
1.00
0.90
Kdg,subcr=0.914
KdTP [-]
0.80
0.70
Kdl=0.734
0.60
Original SRV-ω method
0.50 Original SRV-ζ calculation
Lenzing et al. [1998]-critical
0.40
0 5 10 15 20 25
(b) Kd calculated with ω-method and ζ calculation in industrial SRV/critical regime.
Figure 6.35: Two-phase discharge coefficient versus volumetric quality of air for trans-
parent and industrial valve. Comparison with Lenzing et al. (1998) formula.
this method can be characterized as the most conservative as it has also been reported by
many authors in literature (Tran and Reynolds (2007), Lenzing et al. (1998) and Diener
and Schmidt (2004)).
Results in terms of measured and calculated discharged mass flux versus mass flow
quality obtained in industrial valve are presented in Fig. 6.39. Four different set pressures
are tested; 0.3, 0.6, 0.7 and 0.9 MPa. Calculations with API520 methodology produce
maximum deviations from experimentally determined mass flux of ' 15-20%.
175
Subcritical regime
1.0
Pset=0.3 MPa
0.9
0.8
0.7
0.6
Kd [-]
0.5
0.4
Water
0.3
2%
0.2 5%
10%
0.1
Air-subcritical
0.0
0 1 2 3 4 5 6 7 8
L [mm]
Figure 6.36: Two-phase discharge coefficient versus valve lift in single and two-phase
flow.
P=0.15 MPa−transparent valve
12000
Experimental
Omega
HNE−DS
10000 API 520
8000
Mass flux [kg/m2s]
6000
4000
2000
0 −5 −4 −3 −2 −1 0
10 10 10 10 10 10
Mass quality [−]
Figure 6.37: Experimental and predicted from ω-method, API520 and HNE-DS dis-
charged mass flux at full lift for P set =0.15 MPa in transparent valve.
176
6.7 Correlation for pressure drop in SRV 177
4
x 10 P=0.3 MPa−transparent valve
1.6
Experimental
Omega
1.5 HNE−DS
API 520
1.4
1.3
Mass flux [kg/m2s]
1.2
1.1
0.9
0.8
0.7
0.6 −5 −4 −3 −2
10 10 10 10
Mass quality [−]
Figure 6.38: Experimental and predicted from ω-method, API520 and HNE-DS dis-
charged mass flux at full lift for P set =0.3 MPa in transparent valve.
In industrial SRV, there is a short nozzle upstream the valve therefore a large non-
equilibrium is expected i.e. relaxation time up to the narrowest cross section is low, thus
there is no sufficient time for heat exchange between the two phases. A relatively strong
flow contraction will occur hence a value of α=3/5 should be used for the exponent of
boiling delay factor as it is recommended by the authors Diener and Schmidt. Comparing
results obtained with different methodologies in Fig. 6.39, it is found the most accurate
one is the HNE-DS model with maximum discrepancies less than 5%. Improved precision
of this model is noticed with increasing upstream pressure hence, higher cavitation rates.
Excellent agreement of the model with measured mass fluxes is remarked for the high-
est pressure tested (0.9 MPaG). This conclusion is also reported by Diener and Schmidt
(2004): “...Accordingly, this value is recommended for the calculation of the mass flow
rate through throttling devices with large depressurization rates and short flow lengths...”.
6.7 Correlation for pressure drop in SRV

Since results obtained in singularities cannot be used to model pressure drop in SRV, as it
has been explained in §6.2, pressure drop coefficient is determined for transparent model
of SRV. For two set pressures (1.5 and 3 barg) and several valve openings pressure drop
coefficient is determined from Eq. 6.12.
By fitting experimental results with flow condition parameters, an empirical correla-
tion for ∆P coefficient is established. This coefficient is function of upstream pressure,
valve lift, upstream liquid Re number and volumetric quality (Eq. 6.21).
177
4 P=0.3 MPa−metallic valve 4 P=0.6 MPa−metallic valve

x 10 x 10
2
2.5
Mass flux [kg/m2s]
Mass flux [kg/m2s]

1.5 2
Experimental Experimental
Omega Omega
HNE−DS HNE−DS
API 520 API 520
1 1.5
−5 −4 −3 −2
1 −6 −5 −4 −3
10 10 10 10 10 10 10 10
4 P=0.7 MPa−metallic valve 4 P=0.9 MPa−metallic valve
x 10 x 10
3
Mass flux [kg/m2s]
Mass flux [kg/m2s]

2.5
Experimental Experimental
Omega 2.5 Omega
HNE−DS HNE−DS
2 API 520 API 520
2
1.5
1.5
−6 −5 −4 −3 −6 −5 −4 −3
10 10 10 10 10 10 10 10
Figure 6.39: Calculated versus measured mass flux for different set pressures in industrial
valve.
ζ = 0.961P − 0.247L + 1.322β − 1.675 · 10−6 ReL1 + 5.706 (6.21)

Small variation of upstream pressure is taken into account, however, dependence on
upstream pressure can almost be neglected (regression parameter for P=0.961). Correla-
tion is applied from full lift to 3.5 mm of valve opening; below this value a non-linear
behavior is observed. In total 67 measurement points are used, overall deviation from
measured values is 6.25%, multiple R=0.96 and R2 =0.92.
The relative deviation between predicted with correlation and measured pressure drop
coefficient is plotted in Fig. 6.40. Maximum variation (' 20%) is found for lowest and
highest ζ which correspond to the highest and lowest valve openings (L=7.3 and 3.5 mm)
while for intermediate lift a very good fit from correlation is obtained (≤ 10%).
Summarizing tables with all pressure drop results obtained in transparent SRV are
given in Appendix §K.3.

In this chapter, SRV results are presented. Transparent model and industrial SRV are
compared in terms of discharged flow rate and flowforce applied on valve disk in order to
verify that model simulates correctly the behavior of the actual valve. Flow visualization
178
Correlation-3.5-7.5 mm-67 points
5.00
4.50 10 %
4.00
ζ predicted [-]
3.50 20 %
3.00
2.50
2.00
1.50
1.00
1.00 1.50 2.00 2.50 3.00 3.50 4.00 4.50 5.00
ζ measured [-]
Figure 6.40: Relative discrepancy between predicted and measured pressure drop coeffi-
cient in SRV.
at different parts of the valve reveals the existence of cavitation and provides a qualitative
view of the flow structure. It is proved that only 1-2 % volumetric quality is sufficient
to obtain a chaotic flow regime downstream the valve. This leads to conclusion that
none from the previous correlations deduced for geometrical singularities can be used to
evaluate pressure drop for the valve as a series of geometrical accidents (convergence and
sudden expansion). Additionally, optical probe is used to acquire void fraction and bubble
diameter and velocity upstream and downstream the valve.
Flowforce is measured with force sensor for air and water flow. Characteristic 3D
graphs F=f(P,L) are plotted for both cases and it is revealed that for water flow, force
follows an asymptotically increasing trend with valve lift while for air and same con-
ditions, disk force is relatively high from the lowest lifts (pop-action of the valve) and
increases gradually further on until full lift. Experimental results indicate that compress-
ible and incompressible flowforces in SRV are inversed above a certain value of valve lift.
This value varies with critical pressure ratio, therefore is linked to the position at which
chocked flow occurs during air valve operation. In two-phase flow, for volumetric qual-
ity of air β=20%, pure compressible flow behavior, in terms of flowforce, is remarked at
full lift. Moreover, three-dimensional plots of flowforce-valve opening-set pressure are
deduced to demonstrate the dependency of spring force upon pressure at different valve
lifts.
Simplified axisymmetric 2D numerical simulations are performed and useful conclu-
sions on flow behavior and interaction between valve disk-disk seat and valve body are ex-
tracted. Discharged reflected jet is affected by recirculation regions and sonic line position
(in compressible flow) explaining the flowforce inverse behavior between air and water.
Moreover, comparison between experimental results and CFD computations demonstrates
that the latter provides reasonable prediction of flowforce both in air and water flow with
179
small deviation between tests and computations observed for lowest lift due to experi-
mental uncertainty on adjustment ring location. Influence of adjustment ring location for
the smallest openings is pointed out by quantifying the flowforce variation due to static
pressure gradient at the “sensible” disk-ring area.
Pressure drop coefficient of SRV is calculated in various valve openings and pressure
conditions in single- and two-phase flow. Three dimensional plot of pressure drop coef-
ficient, quality of air and valve lift indicates that the opening characteristics of SRV in
two-phase mixture, even for low qualities, approach more to compressible flow behavior.
Empirical correlation obtained by data fitting to predict pressure drop function of affecting
parameters is proposed and is evaluated against experimental points. Flow rate through
the valve at various openings and set pressures are measured and plotted versus mass
quality. The aforementioned results are also compared with predictions obtained from
several two-phase flow methodologies (HEM, HNE-DS, API 520) with the aim of verify-
ing validity and accuracy range of each model. Therefore, two-phase discharge coefficient
for the specific valve studied is determined using the different modeling approaches. The
HNE-DS methodology is proved to be the most precise among the others.
180
Part V
Conclusions and future

recommendations
7 Discussion
The goal of this thesis was to investigate the behavior of safety relief valve under various
two-phase flow scenarios. As a first step, a series of simple geometrical discontinuities are
tested with the aim of modeling SRV as a sequence of singularities (e.g. contraction and
sudden expansion). Information on the influence of these geometries on pressure drop and
mass flow rate is extracted. From experimental results, pressure drop correlations with no
change of flow regime due to singularity (uniform bubbly dispersed) are deduced.
As a second step, with the purpose of simulating the industrial use of SRV, air is
injected upstream a selected valve in order to reproduce a uniform two-phase air-water
flow (in reality it is usually vapor-liquid) mixture of various qualities. Static behavior
of this specific type of valve (1 1/2" G 3" ) is studied and identical transparent model of
this valve is built and investigated allowing visualization of cavitating phenomena and
injected two-phase flow. From visual observation, it is clear that even for the lowest up-
stream qualities, a chaotic flow regime is produced downstream the valve and therefore no
use of previously obtained correlations on singularities is feasible. Hence, pressure drop
correlation is established using the results on transparent model of the valve. The accu-
racy of different existing methodologies predicting mass flux through throttling device
is tested for non-flashing and flashing conditions (ω-method, HNE-DS method, API520
recommended practice).
Flowforce characterization in compressible, incompressible and two-phase flow con-
ditions of SRV is carried out in order to obtain information on valve opening character-
istics. Numerical simulations with commercial CFD code in simplified axisymmetric 2D
geometry are performed as well to give a better insight of the nature of the flow especially
for the lowest lifts.
7.1 Conclusions
At the first part of the thesis, the common concepts and terminology used in the next
chapters are reported. The main single- and two-phase flow parameters used in singu-
larities and safety relief valve studies are introduced. Several flow patterns in horizontal
and vertical straight pipe flow are discriminated since major influence of flow topology
on pressure drop is expected; inlet and outlet flow conditions are crucial for study of sin-
gularity (horizontal flow) and SRV (horizontal/vertical flow). Furthermore, the various
types of safety valves that can be found in the industry are classified and the vocabulary
used during valve operations is reported. Details on the working principle of SRV are
given and the role of adjusting ring, valve disk and nozzle geometry are explained.
Additionally, analytical derivation and limitations-assumptions of the most important
184 7 Discussion
methodologies used to predict the mass flux through nozzles in two-phase flows, i.e.
the Homogeneous Equilibrium model (HEM), the ω method, the HNE-DS method and
API520 model, are provided. For HEM, homogeneous isentropic and reversible adiabatic
flow is assumed. The ω method is an explicit solution of HEM making the additional
hypotheses of mechanical and thermal equilibrium, ideal gas behavior, that the latent heat
of vaporization of the liquid is constant throughout the nozzle and that vapor pressure and
temperature follow the Clapeyron equation. HNE-DS method is an extension of ω model
which is valid both for equilibrium and non-equilibrium conditions, taking into account
the boiling delay between the two phases. The proposed API520 method is based on
Wallis (1969) drift-flux separated flow model.
Several experimental facilities for the study of geometrical accidents and safety relief
valve are used during this project. Pressure measurements are performed in small (LUCY
II) and large scale (AGATHE) testing smooth and sudden contraction and expansion sec-
tions. Volumetric qualities up to 40% are reached with maximum water flow rate of '4.5
l/s. Additionally, visualization is carried out in selected convergence geometry. For the
same case, optical probe measurements are performed at 6 diameters upstream and down-
stream the geometry to determine local void fraction, bubble diameter and velocity. Safety
relief valve of type API 1 1/2" G 3" is selected and tested in air, water and two-phase flow.
Transparent model of this valve with same inlet and outlet dimensions and small discrep-
ancies from the core and disk of industrial valve is constructed and studied (AGATHE
II). However, important difference between the two valves is found in upstream nozzle
whose length for transparent model is twice as long as in industrial valve. Same transpar-
ent block is used to measure upstream and downstream profiles (void fraction and bubble
velocity) with optical probe and to perform flow visualization in the valve (LUCY III).
Expansion singularities are tested and pressure evolution plots across several singu-
larities are demonstrated and singular pressure change contribution is presumed. Pressure
drop coefficients are extracted and compared for each case. A comparison between litera-
ture proposed methodology and experimental measurements reveals satisfactory agree-
ment with literature models for single and two-phase flows in sudden expansion ge-
ometries. Additionally, a new pressure drop correlation is extracted from experimental
database for each singularity. Flow visualization is performed in divergent section with
high-speed camera and different flow patterns are identified for various flow conditions.
Resulting flow regime maps are established. Optical probe measurements are performed
in the selected case of divergence section of 9 ◦ angle and σ=0.64. The results obtained
are exemplified and void fraction profiles are plotted. The flow structure is compared with
flow visualization for the same conditions. The influence of section expansion on bubble
size distribution is investigated upstream and downstream the singularity.
Convergence singularities are also studied; smooth and sudden geometries of differ-
ent opening angles and surface are ratios are considered. Static pressure evolution plots
are presented. Pressure decreasing regularly upstream the contraction is slightly affected
close to the singularity (1-2 diameters) followed by a sudden pressure drop step and the
influence of vena contracta (strong pressure peak) for maximum 3 upstream diameters
after the singularity. Pressure drop coefficient is deduced for each case. For two-phase
mixtures, influence of liquid Re number is noticed above β=7% for abrupt contraction
while for smooth convergence, liquid Re number affects significantly the pressure drop
even for the smallest volumetric qualities tested. This is possibly due to the unsufficiently
184
7.1 Conclusions 185
good mixing of the two phases for the lateral case. Finally, experimental pressure drop
correlations function of volumetric quality, upstream liquid Re number are extracted.
Industrial Safety Relief Valve and transparent model are compared in terms of dis-
charged flow rate and flowforce applied on valve disk in order to verify that model simu-
lates correctly the behavior of the original valve. Flow visualization at different parts of
the valve reveals the existence of cavitation and provides a qualitative view of the flow
structure. It is proved that only 1-2 % volumetric quality is sufficient to obtain a chaotic
flow regime downstream the valve. This leads to conclusion that none from the previous
correlations deduced for geometrical singularities can be used to evaluate pressure drop
for the valve as a series of geometrical accidents (convergence and sudden expansion).
Additionally, optical probe is used to acquire void fraction, bubble diameter and velocity
upstream and downstream the valve.
Flowforce is measured with force sensor for air and water flow. Characteristic 3D
graphs F=f(P,L) are plotted for both cases and it is revealed that for water flow, force fol-
lows an asymptotically increasing trend with valve lift while for air and same conditions,
disk force is relatively high from the lowest lifts (pop-action of the valve) and increases
gradually further on until full lift. Experimental results indicate that compressible and in-
compressible flowforces in SRV are inversed above a certain value of valve lift. This value
varies with critical pressure ratio, therefore is linked to the position at which chocked flow
occurs during air valve operation. In two-phase flow, for volumetric quality of air β=20%,
pure compressible flow behavior, in terms of flowforce, is remarked at full lift.
Relatively simple axisymmetric 2D numerical simulations are performed and useful
conclusions on flow behavior and interaction between valve disk-disk seat and valve body
are extracted. Discharged reflected jet is affected by recirculation regions and sonic line
position (in compressible flow) explaining the flowforce inverse behavior between air
and water. Moreover, comparison between experimental results and CFD computations
demonstrates that the latter provides very good prediction of flowforce in air and water
with a slight deviation of flowforces in latter conditions, possibly related to experimen-
tal uncertainty on adjustment ring location. Influence of adjustment ring location for the
smallest openings is pointed out by quantifying the flowforce variation due to static pres-
sure gradient at the “sensible” disk-ring area.
Pressure drop coefficient of SRV is calculated in various valve openings and pressure
conditions in single- and two-phase flow. Three dimensional plot of pressure drop coef-
ficient, quality of air and valve lift indicates that the opening characteristics of SRV in
two-phase mixture, even for low qualities, approach to compressible flow behavior. Em-
pirical correlation obtained by data fitting to predict pressure drop function of affecting
parameters is proposed and is evaluated against experimental points. Flow rate through
the valve at various openings and set pressures are measured and plotted versus mass
quality. The aforementioned results are also compared with predictions obtained from
several two-phase flow methodologies (HEM, HNE-DS, API 520) with the aim of verify-
ing validity and accuracy range of each model. Therefore, two-phase discharge coefficient
for the specific valve studied is determined using the different modeling approaches. The
HNE-DS methodology is proved to be the most precise among the others. Finally, we con-
clude that air-water flow with cavitation simulated the behavior of flashing liquid and for
these specific conditions it can be modeled with modified method of Diener and Schmidt
(CF-HNE-DS) although further validation of this observation is required.
185
186 7 Discussion
7.2 Future works-recommendations

Possible future steps are proposed hereunder:
• Perform tests at higher air qualities (> 50% up to pure gas flow) in order to verify
the validity of the conclusions extracted from this thesis.
• Carry out experiments on different types of valves (bigger valves); other kinds of
orifices (such as J orifice). Hence, comparison with the results of flowforce and
pressure drop obtained in orifice G will be feasible. Confirm the accuracy of differ-
ent sizing models for other category of valve.
• Use two-phase flow mixtures that simulate better than air-water the industrial ap-
plication (vapor-liquid flow).
• Measure and locally characterize cavitation bubbles produced for liquid flow. The
latter is very important for the determination of discharge coefficient. Safety Relief
Valve manufacturers consider, when establishing Kd , pure liquid flow (cavitation is
not taken into account), hence incompressible flow which is a doubtful hypothesis.
• Analysis on dynamic operation of the valve (use the valve with the spring). Compar-
ison with static flow results and verification of any potential discrepancies between
dynamic and static behavior of safety relief valve will be possible.
186
Bibliography
AFIR-CETIM. Guide de la Robinetterie Industrielle. Technical report, Association

Française des Industries de la Robinetterie - Centre Technique des Industries Mé-
caniques, 1999.
Ahmed, W. H., Ching, C. Y., and Shoukri, M. Pressure recovery of two-phase flow across
sudden expansions. Int. J. Multiphase Flow, 33:575–594, 2007.
Ahmed, W. H., Ching, C. Y., and Shoukri, M. Development of two-phase flow down-
stream of a horizontal sudden expansion. Int. J. Heat and Fluid Flow, 29:194–206,
2008.
Aloui, F. and Souhar, M. Experimental study of two-phase bubbly flow in a flat duct
symmetric sudden expansion-Part 1: visualization, pressure and void fraction. Int. J.
Multiphase Flow, 22:651–665, 1996.
Aloui, F., Doubliez, L., Legrand, J., and Souhar, M. Bubbly flow in an axisymmetric
sudden expansion: Pressure drop, void fraction, wall shear stress, bubble velocities and
sizes. Experimental Thermal and Fluid Science, 19:118–130, 1999.
Alves, G. E. Cocurrent liquid-gas flow in a pipeline contactor. Chem. Eng. Prog., 50:
449–456, 1954.
API 526. Flanged Steel Pressure Relief Valves. American Petroleum Institute, 5th edition,
2002.
API RP 520. Sizing, Selection and Installation of Pressure-Relieving Devices in Refiner-

ies, Part 1-Sizing and Selection. American Petroleum Institute Recommended Practice
520, 7th edition, 2000.
Armand, A. A. and Treschev, G. G. Investigation of the resistance during the movement

of steam-water mixtures in heated boiler pipe at high pressures. AERE Lib/Trans., 81,
1959.
Arts, T., Boerrigter, H., Buchlin, J. M., Carbonnaro, M., Dénos, R., Degrez, G.,
Fletcher, D., Olivari, D., Riethmuller, M. L., and den Braembussche, R. A. V. Mea-
surement Techniques in Fluid Dynamics, an Introduction. von Karman Institute for
Fluid Dynamics, St. Genesius-Rode, Belgium, 2nd revised edition, 2001.
ASME VIII-Div.1. Boiler and Pressure Vessel Code, Section VIII, Division 1: Rules for
Construction of Pressure Vessels. American Society of Mechanical Engineers, 2010.
188 Bibliography
Bacharoudis, E., Rambaud, P., and Kourakos, V. Numerical modeling of bubbly flows
in divergent-convergent channels. VKI PR 2008-49, von Karman Institute for Fluid
Dynamics, St. Genesius-Rode, Belgium, 2008.
Baker, O. Simultaneous flow of oil and gas. Oil Gas J., 53:185, 1954.
Bankoff, S. G. A variable density single fluid model for two-phase flow with particular
reference to steam-water flow. J. Heat transfer, 82:265, 1960.
Barnea, D., Shoham, O., and Taitel, Y. Flow pattern transition for vertical downward
two-phase flow. Chemical Engineering Science, 37:741–744, 1982.
Baroczy, C. J. A systematic correlation fot two-phase pressure drop. Chem. Eng. Prog.
Symp. Ser., 62:232–249, 1966.
Beattie, D. R. H. A note on the calculation of two-phase pressure losses. Nucl. Eng. Des.,
25:395–402, 1973.
Bell, K. J., Taborek, J., and Fenoglio, F. Interpretation of horizontal in-tube condensation
heat transfer correlations with a two-phase flow regime map. Chem. Eng. Prog. Symp.
Ser., 66:150–163, 1970.
Bereznai, G. T. Nuclear power plant systems and operation. Technical Report Revision
4, School of Energy Systems and Nuclear Science University of Ontario Institute of
Technology, Oshawa, Ontario, 2005.
Bertola, V. Optical probe visualization of air-water flow structure through sudden area
contraction. Experiments in Fluids, 32:481–486, 2002.
Bertola, V. The structure of gas-liquid flow in horizontal pipe with abrupt area contraction.
Experimental Thermal and Fluid Science, 28:505–512, 2004.
Beune, A. Analysis of high pressure safety valve. PhD thesis, Eindhoven University of
Technology, Eindhoven, 2009.
Blasius, P. R. H. Das aehnlichkeitsgesetz bei reibungsvorgangen in flussigkeiten.

Forschungsheft, 131:1–41, 1913.
Boccardi, G., Bubbico, R., Celata, G. P., and Mazzarotta, B. Two-phase flow through
pressure safety valves. Experimental investigation and model prediction. Chemical
Engineering Science, 60:5284–5293, 2005.
Brennen, C. E. Fundamentals of multiphase flows. Cambridge University Press, Califor-

nia Institute of Technology, Pasadena, California, 2005.
Chabane, S., Plumejault, S., Pierrat, D., Couzinet, A., and Bayart, M. Vibration and chat-
tering of conventional safety relief valve under built up back pressure. In 3rd IAHR
International Meeting of the WorkGroup on Cavitation and Dynamic Problems in Hy-
draulic Machinery and Systems, October 14-16, 2009, Brno, Czech Republic, pages
281–294. International Association of Hydro-Environment Engineering and Research,
2009.
188
Bibliography 189
Cheremisinoff, N. P. and Gupta, R. Handbook of Fluids in Motion. Ann Arbor Science

Book, 1983.
Chisholm, D. Theoretical aspects of pressure changes at changes of section during steam-

water flow. NEL rept. 418, 1969.
Chisholm, D. Prediction of pressure drop at pipe fittings during two-phase flow. In Proc.
13th Int. Conf. Inst. Refrig. Congr., Washington, USA, 1971.
Chisholm, D. Friction during the flow of two-phase mixtures in smooth tubes and chan-
nels. Technical Report NEL rept., 1976.
Colebrook, C. F. and White, C. M. Experiments with fluid friction in roughened pipes.

Proceedings of the Royal Society of London. Series A, Mathematical and Physical Sci-
ences, 161(906):367–381, 1937.
Collier, J. G. Convective boiling and condesensation. McGraw-Hill, London, 1972.
Collier, J. G. and Thome, J. R. Convective boiling and condesensation. Clarendon Press,

Oxford, 3rd edition, 1994.
Comolet, R. Mécanique expérimentale des fluides.-Tome II : Dynamique des fluides réels,

turbomachines. Masson, 4ème edition, 1963.
Corbin, F. and Chabane, S. Mesure des efforts sur une soupape de sûreté. Rapport CETIM
CET0040650/6G1/a, Centre Technique des Industries Mécaniques, Techniques des flu-
ides et des écoulements, 74 route de la Jonelière BP 82617, 44326 Nantes Cedex 3,
2009.
Corbin, F., Pozzoli, R., and François, P. Essais de soupapes-Banc eau. Documents Qualité
CETIM T-8600-a, Centre Technique des Industries Mécaniques, Techniques des fluides
et des écoulements, 74 route de la Jonelière BP 82617, 44326 Nantes Cedex 3, 2009a.
Corbin, F., Pozzoli, R., and François, P. Appareils de mesure-étalonnage incertitude de

mesure. Documents Qualité CETIM T-5389 b, Centre Technique des Industries Mé-
caniques, Techniques des fluides et des écoulements, 74 route de la Jonelière BP 82617,
44326 Nantes Cedex 3, 2009b.
Darby, R., Meiller, P. R., and Stockton, J. R. Select the best model for two-phase relief
sizing. a variety of methods exist for sizing valves, but not all give the best predictions
for certain conditions. Chem. Eng. Prog., 97(5):56–65, 2001.
Darcy, H. Recherches experimentales relatives au mouvement de l’eau dans les tuyaux.

Mallet-Bachelier, Paris, 1857. 268 pages and atlas (in French).
Dehaeck, S., Planquart, P., and Riethmuller, M. L. Two-phase flow in continuous cast-
ing. VKI PR 2003-08, von Karman Institute for Fluid Dynamics, St. Genesius-Rode,
Belgium, 2003.
189
190 Bibliography
Delgado-Tardáguila, R., Buchlin, J.-M., Rambaud, P., and Kourakos, V. Study of geomet-
rical singularities in liquid and gas-liquid flow. VKI PR 2008-06, von Karman Institute
for Fluid Dynamics, St. Genesius-Rode, Belgium, 2008.
Delhaye, J. M., Giot, M., and Riethmuller, M. L. Thermohydraulics of Two-Phase Systems
for Industrial Design and Nuclear Engineering. Hemisphere Pub. Corp., McGraw-Hill
Book Company, New York, 1980.
Deniz, E., Buchlin, J.-M., and Kourakos, V. Experimental study of bubbly flow in singular
geometries. VKI PR 2009-09, von Karman Institute for Fluid Dynamics, St. Genesius-
Rode, Belgium, 2009.
Descartes, R. La dioptrique. 1673.
Diener, R. and Schmidt, J. Sizing of throttling device for gas/liquid two-phase flow. Part
1: Safety valves. Process Safety Progress, 23(4):335–344, 2004.
Diener, R. and Schmidt, J. Sizing of throttling device for gas/liquid two-phase flow. Part
2: Control valves, orifices, and nozzles. Process Safety Progress, 24(1):29–37, 2005.
Dix, G. E. Vapor void fractions for forced convection with subcooled boiling at low flow
rates. Technical Report NEDO-10491, General Electric Company, 1971.
Dossena, V., Gaetani, P., Marinoni, F., and Osnaghi, C. On the influence of back pressure
and size on the performance of safety valves. In PVP-Vol. 447 Piping and component
analysis and Diagnostics, pages 35–42. ASME, 2002.
Dossena, V., Marinoni, F., Di Vicenzo, S., Boccazzi, A., and Sala, R. High pressure
fluctuations induced by safety valves operating with liquids at very low lift. In PVP-
Vol. 488 Risk and Reliability and Evaluation of Components and Machinery, pages
95–10. ASME, 2004.
EDSU. The frictional component of pressure gradient for two-phase gas or vapor/liquid
flow through straight pipes. In Eng.Sci., Data Unit (EDSU), London, 1976.
Endress+Hauser. Application de la directive européenne sur les equipements sous pres-
sion (DESP)-Directive 97/23/CE. http://www.fr.endress.com/eh/sc/europe/
fr/fr/home.nsf/?Open&DirectURL=7CE49EA9A4C398ADC125745E004BD164,
2010.
Epstein, M., Henry, R. E., Midvidy, W., and Pauls, R. One-dimensional modeling of two-
phase jet expansion and impingement. In In 2nd International Topical Meeting Nuclear
Reactor Thermal•Hydraulics, American Nuclear Society, Santa Barbara, 1983.
Fernandes, V., Kourakos, V., and Buchlin, J.-M. Investigation of two-phase bubbly flow
in a horizontal sharp bend. VKI PR 2010-08, von Karman Institute for Fluid Dynamics,
St. Genesius-Rode, Belgium, 2010.
Ferrari, J. and Leutwyler, Z. Measurement of the fluid flow load on a globe valve
stem under various cavitation conditions. http://hal.archives-ouvertes.fr/
hal-00413303/fr, 2008.
190
Bibliography 191
Föllmer, B. and Schnettler, A. Challenges in designing api safety relief valves. http:
//www.valve-world.net, oct 2003.
Fossa, M. and Guglielmini, G. Dynamic void fraction measurements in horizontal ducts

with sudden area contraction. Int. J. Heat and Mass Trans., 41:3807–3815, 1998.
François, F., Garnier, J., and Cubizolles, G. A new data acquisition system for binary
random signal application in multiphase flow measurements. Meas. Sci. Technol., 14:
929–942, 2003.
Friedel, L. Mean void fraction and friction pressure drop: comparison of some correla-
tions with experimental data. In European two-phase flow group meeting, Grenoble,
1977.
Ghajar, A. J. Non-boiling heat transfer in gas-liquid flow in pipes–a tutorial. In 10th

Brazilian Congress of Thermal Sciences and Engineering, Rio de Janeiro, RJ, Brazil,
2004.
Golan, L. P. and Stenning, A. H. Two-phase vertical flow maps. Proc. Inst. Mech. Eng.,
184:110–116, 1969.
Goßlau, W. and Weyl, R. Strömungsdruckverluste und reaktionskräfte in rohrleitungen

bei notentspannung durch sicherheitsventile und bertscheiben. Sonderdrucke Techn.
Überw, 30:5–9, 1989.
Govier, G. W. and Aziz, K. The flow of complex mixtures in pipes. Van Nostrand Reinhold,
New York, 1972.
Guglielmini, G., Muzzio, A., and Sotgia, G. The structure of two-phase flow in ducts
with sudden contractions and its effects on the pressure drop. In Experimental Heat
Transfer, Fluid Mechanics and Thermodynamics, pages 1023–1036, 1997.
Gundersen, H. J. and Jensen, E. B. Particle sizes and their distributions estimated from
line- and point-sampled intercepts. including graphical folding. Journal of Microscopy,
131(3):291–310, 1983.
Henry, R. E. and Fauske, H. K. Two-phase critical flow at low qualities. Nucl. Sci. Eng.,
41:79–91, 1970.
Henry, R. E. and Fauske, H. K. The two-phase critical flow of one-component mixtures

in nozzles, orifices, and short tubes. Journal of Heat Transfer, 93:179–187, 1971.
Hewitt, G. F. and Hall-Taylor, N. S. Annlar Two-Phase Flow. Pergamon, Oxford, 1970.
Hewitt, G. F. and Roberts, D. N. Studies of two-phase flow patterns by simultaneous

X-ray and flash photography. In AERE-M 2159, 1969.
Idel’Cik, I. E. Memento des pertes de charge. Editions Eyrolles, 61 Bd Saint-Germain

Paris, 5ième edition, 1986.
191
192 Bibliography
Idsinga, W., Todreas, N. E., and Bowring, R. An assessment of two-phase pressure drop
correlations for steam-water systems. Int. J. Multiphase Flow, 3:401–413, 1977.
Ishii, M. One-dimensional drift-flux model and constitutive equations for relative mo-
tion between phases in various two-phase flow regimes. Technical Report ANL-77-47,
1977.
ISO. Iso-lite software, two-phase flow study-user’s guide. Technical documentation, RBI
Instrumentation et mésure, 2008.
ISO 4126-1. Safety devices for protection against excessive pressure–Part 1: Safety
valves. International Organization for Standardization, 2004.
ISO 4126-10. Safety devices for protection against excessive pressure–Part 10: Sizing of
safety valves for gas/liquid two-phase flow. International Organization for Standardiza-
tion, 2010.
Jannsen, E. and Kervinen, J. A. Two-phase pressure drop across contractions and expan-
sions of water-steam mixture at 600 to 1400 psia. Technical Report Geap 4622-US,
1966.
Joukowsky, N. “über den hydraulischen Stoss in Wasserleitungsrohren”, (“On the hy-
draulic hammer in water supply pipes”. Memoires de l’Academie Imperiale des Sci-
ences de St.-Petersbourg (1900), 9(5):1–71, 1898. (in German).
Korteweg, D. J. “über die Fortpflanzungsgeschwindigkeit des Schalles in elastischen
Rohren” (“On the velocity of propagation of sound in elastic tubes”). Annalen der
Physik und Chemie, 1878.
Kourakos, V., Rambaud, P., and Buchlin, J.-M. Effect of geometrical restriction in gas-
liquid flow. VKI PR 2007-13, von Karman Institute for Fluid Dynamics, St. Genesius-
Rode, Belgium, 2007.
Kourakos, V., Chabane, S., Rambaud, P., and Buchlin, J.-M. Hydrodynamic forces, pres-
sure and mass flux in two-phase air-water flow through transparent safety valve model.
In ASME 2010 Pressure Vessels and Piping Conference, PVP 2010, Washington, USA,
2010a.
Kourakos, V., Deniz, E., Rambaud, P., Chabane, S., and Buchlin, J.-M. Investigation of
the void fraction, bubble size and velocity in divergence geometry. In 7th International
Conference of Multiphase Flow ICMF 2010, Tampa, FL, USA, 2010b.
Kourakos, V. G., Rambaud, P., Chabane, S., and Buchlin, J.-M. Two-phase flow modeling
within expansion and contraction singularities. In 5th Int. Conf. on Computational and
Experimental Methods in Multiphase and Complex Flow, pages 27–40, Computational
Methods in Multiphase Flow V, Vol. 63, WITpress2009, 2009.
Kourakos, V. G., Rambaud, P., Chabane, S., and Buchlin, J.-M. Modelling of pressure
drop in two-phase flow within expansion geometries. In 6th International Symposium
on Multiphase Flow, Heat mass transfer and Energy Conversion, volume 1207, pages
802–808, AIP conf. Proc., 2010c.
192
Bibliography 193
Lenzing, T. and Friedel, L. Full lift safety valve air/water and steam/water critical mass
flow rates. 11th Mtg ISO/TC185/WG1, 1996.
Lenzing, T. and Friedel, L. Vorhersage des maximalen Massendurchsatzes von Vollhub-
sicherheitsventilen bei Zweiphasenströmung. TÜ 39, 6, 1998.
Lenzing, T., Friedel, L., Cremers, J., and Alhusein, M. Prediction of the maximum full
lift safety valve two-phase flow capacity. Journal of Loss Prevention in the Process
Industries, 11:307–321, 1998.
Leung, J. C. Easily size relief devices and piping for two-phase flow. Chem. Eng. Prog.,
92(12):28–50, 1996.
Lockhart, R. W. and Martinelli, R. C. Proposed correlation of data for isothermal two-
phase two-component flow in pipes. Chem. Eng. Prog., 45:39–48, 1949.
Lottes, P. Expansion losses in two-phase flow. Nucl. Sci. Eng., 9:26–31, 1960.
Mandhane, J. M., Gregory, G. A., and Aziz, K. A flow pattern map for gas-liquid flow in
horizontal pipes. Int. J. Multiphase Flow, 1:537–553, 1974.
Martinelli, R. C. and Nelson, D. B. Prediction of pressure drop during forced circulation
boiling of water. Trans. ASME, 79:695–702, 1948.
Moncalvo, D., Friedel, L., Jörgensen, B., and Höhne, T. Sizing of safety valves using
ansys CFX-Flo. Chem. Eng. Technol., 32(2):247–251, 2009.
Moody, L. F. Friction factors in pipe flow. Trans, ASME, 66:671, 1944.
Nastoll, W. Methodik zur abschätzung von kohlenwasser-stoff-freisetzungen mit
zweiphasigen entpannungsströmungen. DGMK-Projekt 248-03, 1985.
National Commission on the BP Deepwater Horizon Oil Spill and Offshore Drilling. Stop-
ping the spill: The five-month effort to kill the Macondo Well. Technical Report Staff
working paper No. 6, 2010.
NRC. Fact sheet on the Three Mile Island accident, U.S. Nuclear Regulatory Com-
mission. http://www.nrc.gov/reading-rm/doc-collections/fact-sheets/
3mile-isle.html, Aug. 2009.
Oertel, H., Erhard, P., Etling, D., Müller, U., Riedel, U., Sreenivasan, K. R., Warnatz, J.,
and Prandtl, L. Prandtl’s essentials of fluid mechanics, volume 158. Springer, Applied
Mathematical Sciences, 3rd edition, 2009.
Oshinowo, T. and Charles, M. E. Vertical two-phase flow, Part 1, flow pattern correlations.
Can. J. Chem. Eng., 52:25–35, 1974.
Paullin, R. L. and Santman, L. D. Report on San Juan Ixhautepec, Mexico LPG Accident.
Technical report, 1985.
PED. Pressure Equipment Directive-PED. 97/23/EC, European Commission - Enterprise
and Industry, 1997.
193
194 Bibliography
Pietersen, C. M. Analysis of the lpg-disaster in mexico city. Journal of Hazardous Mate-

rials, 20:85–107, 1988.
Schlünder, E. U. Heat exchanger design handbook. Hemisphere Publishing, 79 Madison

Avenue, New York, NY 10016, 1983.
Schmidt, J. Sizing of nozzles, venturies, orifices, control and safety valves for initially
sub-cooled gas/liquid two-phase flow - the HNE-DS method. Forsch Ingenieurwes, 71:
47–58, 2007.
Schmidt, J. Sizing of safety valves for multi-purpose plants according to iso 4126-10.
Forsch Ingenieurwes, 25:181–191, 2011.
Schmidt, J. and Friedel, L. Two-phase pressure change across sudden expansions in duct
areas. Chem. Eng. Comm., 141:175–190, 1996.
Schmidt, J. and Friedel, L. Two-phase pressure drop across sudden contractions in duct
areas. Int. J. Multiphase Flow, 23(2):283–299, 1997.
Schnettler, A. Anforderungen an federsicherheitsventile und auslegung der bauarten im

hinblick auf eine zuverlässige funktion und der einbaubedingungen (federgesetze fur
sicherheitsventile aus kraftkennfeldmessungen). Vortrag. 24.11.1994. Haus der Tech-
nik e.v., Essen, 1994.
Shannak, B. Mixture loss coefficient of safety valves used in nuclear plants. Nucl. Eng.
Des., 239:1779–1788, 2009.
Shannak, B., Friedel, L., and Alhusein, M. Prediction of single- and two-phase flow
contraction through sharp-edged short orifice. Chem. Eng. Technol., 22(10):865–870,
1999.
Song, X. G., Cui, L., and Park, Y. C. Three dimensional CFD analysis of a spring-loaded
pressure relief valve from opening to re-closure. In ASME 2010 Pressure Vessels and
Piping Conference, PVP 2010, Washington, USA, 2010.
SS Engineers & Consultants. Product range-industrial pipe fit-

tings. http://www.indiamart.com/ssengineersandconsultants/
industrial-pumps-pipe-fittings.html, 2010.
Stares, J. A. Control valve cavitation, damage control-Dresser-Masoneilan.

http://www.iceweb.com.au/Valve/Control%20Valves/Masoneilan/
ControlValveCavitation.pdf, 2007.
Taitel, Y. and Dukler, A. E. Model for predicting the flow regime transitions in horizontal
and near horizontal gas-liquid. AIChE J., 22(1):47–55, 1976a.
Taitel, Y. and Dukler, A. E. A model of predicting flow regime transitions in horizontal

and near horizontal gas-liquid flow. AIChE J., 22:47–55, 1976b.
194
Bibliography 195
Taitel, Y. and Dukler, A. E. Flow regime transitions for vertical upward gas-liquid flow:
A preliminary approach through physical modeling. In AIChE 70th Annual Meet., New
York, Session on Fundamental Research in Fluid Mechanics, 1977.
Taylor, J. R. An introduction to uncertainty analysis; the study of uncertainties in physical

measurements. Mill Valley: University Science Books, 1982.
Thom, J. R. S. Prediction of the pressure drop during forced circulation boiling of water.
Int. J. Heat Mass Transfer, 7:709, 1964.
Todreas, N. and Kazami, M. Nuclear Systems I: Thermal Hydraulic Fundamentals, Mas-

sachusetts Inst. of Tech., Cambridge, MA (USA). Hemisphere Publishing, 79 Madison
Avenue, New York, NY 10016 (USA), 2nd edition, 1989.
Tran, Q. K. and Reynolds, M. Sizing of relief valves for two-phase flow in the bayer
process, Kaiser Enginners PTY limited, QV.1 Building 250 St. George”s Terrace,
Perth, Western Australia 6000. http://www.hatch.ca/Light_Metals/Articles/
sizing_relief_valves.pdf, 2007.
Tyco. Pressure relief valve engineering handbook. Technical Publication No. TP-V300,
Tyco Flow Control, 2008.
Validyne. General operating instructions. validyne variable reluctance pressure trans-

ducer. Technical notes, Validyne Engineering, 2005.
Vallet, C., Ferrari, J., Rit, J.-F., and Dehoux, F. Single-phase CFD inside a water safety
valve. In ASME 2010 Pressure Vessels and Piping Conference, PVP 2010, Washington,
USA, 2010.
Velasco, I. L’ écoulement diphasique à travers un élargissement brusque. PhD thesis,

Université Catholique de Louvain, thèse de maitrise, 1975.
Wallis, G. B. One Dimensional Two-Phase Flow. McGraw-Hill, New York, 1969.
Weisbach, J. L. Lehrbuch der Ingenieur-und Maschinen-Mechanik, Vol. 1. Theoretische

Mechanik. Vieweg und Sohn, Braunschweig, 1845. 535 pages (in German).
Weisman, J. Two-phase flow patterns, chapter 15, pages 409–425. Ann Arbor Science
Publ., 1983. eds: N. P. Cheremisinoff and R.Gupta.
Wylie, E. B. and Streeter, V. L. Fluid Transients. McGraw-Hill Book Company, New

York, 1978.
Zuber, N. and Findlay, J. A. Average volumetric concentration in two-phase flow systems.

Trans, ASME J. Heat transfer, 87:453, 1965.
195
Appendices
A Uncertainty analysis
Estimation of the uncertainty on experimental results presented in the previous chapters

is crucial for the further use of the provided database. When calculations are made with
equation giving an exact solution, the error of each measured or estimated quantity is
propagated in the final result. Therefore, in this chapter an attempt to calculate the error
associated to the measurement of each recorded parameter during singularities study is
presented. It is important to point out that, in the present study, all measured quantities
during experiments are based on averaging of the selected number of samples at certain
sampling rate for each factor. Hence, uncertainties are calculated based on these averages
and not individual samples. Good reference for introduction to uncertainty analysis are
given in Taylor (1982) and Arts et al. (2001). The generalized formula that is used in the
next sections to estimate the total uncertainty of calculated parameters (such as pressure
drop coefficient ζ) is given hereunder.
Supposing that the calculated quantity Y is function of i parameters xi (Y=Y(xi )).
Then, according to Taylor’s expansion the uncertainty of variable Y will be given by:
v
t i
X ∂Y 2
δY = δxi2 · (A.1)
1
∂xi
where δxi the uncertainty of parameter xi

The main quantity that had to be estimated concerning singularities study was the
pressure drop coefficient. Since this parameter depends on almost all the other measured
quantities, its uncertainty involves estimation of uncertainty of flow rater, static pressure
measurement etc; the latter are computed in the next paragraphs. Additionally, the signif-
icant uncertainty in the measurements performed with optical probe should be computed.
A.1 Pressure drop coefficient formula

The calculation of pressure drop coefficient (Eq. 4.1 and 4.2) for both contraction and ex-
pansion geometries, taking into account the definition of average velocity of the fluid, can
be expressed function of experimentally measured and calculated parameters (Eq. A.2).
!
−2 · A21 ∆P sing
st + A21 · ρ::
∗
Ut22 − Ut12
::::: ::::::::
ζ= . (A.2)
ρL · Q2L
::
200 A Uncertainty analysis
where the highlighted terms, by a wavy underline, correspond to estimation of the

dependent parameters for which a separate uncertainty should be estimated. However, we
can observe that not all quantities are independent i.e. ρ∗ and Ut ). Thus, in order to be
able to easily calculate the partial derivatives, we develop Eq. A.2 so that pressure drop
coefficient is written function of independent variables and we obtain the following final
form:
2A21 · ∆P sing ρG QG2 QG ρG QG

" # !
st 1
ζ=− + 2 −1 · + + +1 (A.3)
ρL · Q2L A ρL Q2L QL ρL QL
with measured values; ∆P sing

st , QL and QG and the rest being constant (ρL considered
to have negligible variation and ρG known for a given pressure).
The derivativesThe partial
to be derivatives
calculated willfollowing:
are the be:
∂ζ 4 A12 ⋅ ΔPstsing ⎛ 1 ⎞ ⎛ 2ρ Q Q ρ Q ⎞
=− + ⎜1 − 2 ⎟ ⎜ G 3G + G2 + G G2 ⎟
∂QL ρ LQL3
⎝ A ⎠ ⎝ ρ LQL QL ρ LQL ⎠
∂ζ ⎛ 1 ⎞ ⎛ 2ρ Q 1 ρ ⎞
= ⎜ 2 − 1⎟ ⎜ G 2G + + G ⎟
∂QG ⎝ A2 ⎠ ⎝ ρ L QL QL ρ L QL ⎠
∂ζ 2 A12
= −
∂ΔPstsing ρ LQL2 (A.4)
Therefore, if the uncertainties of static pressure change, water and air flow rates are
calculated the final uncertainty of ζ can be calculated by combination of Eq A.1 Eq A.4.
Uncertainty in water and air flow rate

As it has been already mentioned in §3.1.2, the uncertainty related to water measure-
ment with electromagnetic flowmeter is 0.5-1.10% for as it has been calculated in curve
A3 - reference document Corbin et al. (2009b). For calculation of the uncertainty, we
will choose the worst scenario which has been in the sudden expansion measurement of
σ=0.43 measuring at 30% of the FS which corresponds to a relative uncertainty of 0.8%
in water flow rate measuement, hence δQL =0.8%.
Concerning the measurement of air flow rate, in the case of large scale experiments
(AGATHE) the air mass flow rate is measured contrary to small scale (LUCY II) for which
directly the volume flow rate is recorded. The mass flow meter working in a range from
0-12 g/s (0-10 V) with an accuracy of 0.26-0.44 (100-20% FS). Since our measurements
where restricted to maximum 5 g/s, except for one case of nitrogen use for SRV, the
apparatus measurement range is fixed to 0-5 g/s for 0-10 Volts, increasing in this way
significantly the accuracy of the measurement. Estimating the lowest mass flow rate tested
during the whole measurement campaign and transforming it to volume flow rate, a final
relative uncertainty of the gas flow rate equal to δQG =0.44% is approximated.
200
A.1 Pressure drop coefficient formula 201
Uncertainty in static pressure change

For the determination of static pressure change, the pressure between two points in the
pipe are measured with differential pressure transducers. According to documentation by
Corbin et al. (2009b), the transducer is calibrated at B3 curve therefore measuring with a
precision 0.35-0.75 % for 100-20% of the full scale. Several devices have been used to
cover the whole area of pressure map managing to always use them at minimum 40% of
FS. This leads to a final value of relative uncertainty in ∆P equal to 0.5% according to the
curve δ-FS; ∆P sing
st =0.5%.
Final uncertainty in pressure drop coefficient ζ

Summing up the results obtained in the previous three sections, we can use Eq. A.1 and
Eq. A.4 to obtain the total pressure drop coefficient uncertainty. Hence, the following
final matrix is built (Table A.1). We can remark from this table that the uncertainties are
relatively low (of the order of 1-3%).
sing
Geometry QL [m3/h] QG [m3/h] ∆Pst [mbar] δQL [m3/h] δQG [m3/h] δ∆Pstsing [mbar] δζ [-]
σ=0.43, α=90° 7.08 0 76 0.008 0.0044 0.005 0.653
7.02 0.272 78 0.008 0.0044 0.005 0.661
5 0.167 40.6 0.008 0.0044 0.005 0.926
5.01 0.358 41.6 0.008 0.0044 0.005 0.932
5.01 1.54 37.9 0.008 0.0044 0.005 1.06
σ=0.43, α=8.2° 3.01 0 17.4 0.008 0.0044 0.005 1.54
3.01 0.0451 16.4 0.008 0.0044 0.005 1.54
3.02 0.105 16 0.008 0.0044 0.005 1.54
3.01 0.162 15.7 0.008 0.0044 0.005 1.55
3.02 0.221 15.4 0.008 0.0044 0.005 1.55
σ=2.34, α=90° 7.39 0 160 0.008 0.0044 0.005 3.43
7.4 0.209 167 0.008 0.0044 0.005 3.43
7.39 0.354 172 0.008 0.0044 0.005 3.44
7.39 0.537 181 0.008 0.0044 0.005 3.46
7.4 0.787 194 0.008 0.0044 0.005 3.49
7.4 1.09 209 0.008 0.0044 0.005 3.55
σ=2.34, α=14.8° 7.39 1.28 208 0.008 0.0044 0.005 3.6
8.06 0 154 0.008 0.0044 0.005 3.14
8.58 0.271 181 0.008 0.0044 0.005 2.96
8.6 0.455 185 0.008 0.0044 0.005 2.96
8.61 0.666 190 0.008 0.0044 0.005 2.97
Uncertainty in optical probe measurements

An overall estimation of total uncertainty on void fraction measurements has been given
to §4.3 and it was of the order of 10-15 %.
201
B Design of experimental facilities
Detailed drawings of the test section for optical probe in LUCY II are presented in Fig.
B.1. The 4 measurement locations on upstream position and the exact dimensions of
divergence-convergence section are indicated.
Furthermore, the plan of the transparent SRV (API 1 1/2" G 3" ) test section (AGATHE
II) is given in Fig. B.2. Pressure taps can be distinguished upstream, downstream and in
the core of the transparent model of the valve.
Plans of the modified transparent SRV facility for optical probe measurements (LUCY
III) is presented in Fig. B.3. One measurement position is placed upstream the valve while
two are located downstream the model.
The sketch of the force sensor mounted in the original safety relief valve is shown in
Fig. B.4. A system that allows static variation of the valve opening designed.
Optical probe
204
204
Figure B.1: Detailed optical probe test section.
A-A
670
Ecrou de réglage de
l'ouverture de la soupape
B5
Course soupape 11
P
s
Bride PN16 DN80

Fixation par 8 vis M16 22 Bague d'arrêt B2
Butée de réglage de la
course de la soupape
Jeu 3
B1
91
200
B3
Jeu 2
6 Vis H M12x50 6 rondelles larges M10
26
287
B4
Bride PN10 DN40

Fixation par 4 vis H M12x35
205
205
Figure B.2: Transparent SRV test section AGATHE II.

206 B Design of experimental facilities
0,10
3 x Ø14 +- 0,00
P
80
0
P9
270
90
59,5
119
Figure B.3: LUCY III test section drawing.
206
B Design of experimental facilities 207
Plan of force sensor mounting
Figure B.4: Force sensor mounted in safety valve WEIR of type 1 1/2 G 3 .
00 00
207
C Disasters caused by PRV failures
Three Mile Island accident (1979)

The main cause of release of radioactivity in the Three Mile Island accident in 1979 was a
Pilot-Operated Relief Valve (PORV) on the primary loop which stuck in the open position.
This caused the overflow tank into which it drained to rupture and released large amounts
of radioactive cooling water into the containment building.
The accident began at 4 a.m. on Wednesday, March 28, 1979, with failures in the
non-nuclear secondary system, followed by a stuck-open PORV in the primary system,
which allowed large amounts of reactor coolant to escape. The mechanical failures were
compounded by the initial failure of plant operators to recognize the situation as a LOCA
due to inadequate training and human factors, such as industrial design errors relating to
ambiguous control room indicators in the power plant’s user interface (Fig. C.1).
Due to the loss of heat removal from the primary loop and the failure of the auxiliary
system to activate, the primary side pressure began to increase, triggering the PORV at the
top of the pressurizer to open automatically. The PORV should have closed again when
the excess pressure had been released and electric power to the solenoid of the pilot was
automatically cut, but instead the main relief valve stuck open due to a mechanical fault.
The open valve permitted coolant water to escape from the primary system, and was the
principal mechanical cause of the crisis that followed.
Critical human factors problems were revealed in the investigation about the industrial
design of the reactor control system’s user interface. A lamp in the control room, designed
to illuminate when electric power was applied to the solenoid that operated the pilot valve
of the PORV, went out, as intended, when the power was removed. This was incorrectly
interpreted by the operators as meaning that the main relief valve was closed, when in
reality it only indicated that power had been removed from the solenoid, not the actual
position of the pilot valve or the main relief valve. Because this indicator was not designed
to unambiguously indicate the actual position of the main relief valve, the operators did
not correctly diagnose the problem for several hours.
The operators had not been trained to understand the ambiguous nature of the PORV
indicator and look for alternative confirmation that the main relief valve was closed. There
was a temperature indicator downstream of the PORV in the tail pipe between the PORV
and the pressurizer that could have told them the valve was stuck open, by showing that the
temperature in the tail pipe remained high after the PORV should have, and was assumed
to have, shut, but this temperature indicator was not part of the “safety grade” suite of
indicators designed to be used after an incident, and the operators had not been trained to
use it. Its location on the back of the desk also meant that it was effectively out of sight of
210 C Disasters caused by PRV failures
the operators (NRC (2009)).
Figure C.1: Three Mile Island nuclear power plant.

Three Mile Island nuclear plant
Mexico City (1984)

An example of accident in Mexico City caused by the lack of sufficient number of safety
valves is demonstrated in Fig. C.2. A Boiling Liquid Expanding Vapor Explosion (BLEVE)
at a Liquefied Petroleum Gas (LPG) terminal near Mexico City resulted in 650 deaths and
over 6.400 injuries. Damage to the plant was estimated at $ 31.3 million. The probable
cause of the accident was over-pressuring of the LPG facilities from the pressure of the
pipeline supplying the plant, as it is reported by B. Olson, R. L Paullin and L. D. Santman
(Paullin and Santman (1985)). More precisely, it is stated about the possible cause of
the accident: “...The probable rapid release of a large volume of gas discharged prior to
failure through the safety valves of the over-pressured tank and three others which may
have been being filled at the same time, coupled with the high probability that pumping
continued at full flow for an additional hour after the explosion, explain the massiveness
of the spill and why no alarm or warning was received by the community...”.
More information about this disaster can be found in Pietersen (1988).
Deepwater Horizon oil spill (2010)

Another very severe accident in terms of environmental and financial consequences has
recently occurred in the Gulf of Mexico; the Deepwater Horizon oil spill.
The Deepwater Horizon oil spill, also referred to as the BP (British Petroleum) oil
spill, is an oil spill in the Gulf of Mexico which flowed for three months in 2010. It is
the largest accidental marine oil spill in the history of the petroleum industry. The spill
stemmed from a sea-floor oil gusher that resulted from the April 20, 2010 Deepwater
Horizon drilling rig explosion. The explosion killed 11 platform workers and injured 17
210
C Disasters caused by PRV failures 211
Figure C.2: Accident mainly caused by missing safety valves (Mexico city 1984).
others. On July 15, the leak was stopped by capping the gushing wellhead after releasing
about 4.9 million barrels (780·103 m3 ), or 185 million gallons of crude oil. It is believed
that the daily flow rate diminished over time, starting at about 62,000 barrels per day
(9,900 m3 /d) and decreasing as the reservoir of hydrocarbons feeding the gusher was
gradually depleted. On September 19, the relief well process was successfully completed.
According to BP’s internal investigation, the reason of the explosion was the leakage
of Methane gas from the well that ignited after expanding out of the well. One more
reason that is coming into picture is the cost cutting by BP on maintenance activities that
ent caused the explosion.
The safety valve that failed to prevent the Deepwater Horizon rig exploding in the Gulf
of Mexico was wrongly plumbed, according to a senior BP official (National Commission
on the BP Deepwater Horizon Oil Spill and Offshore Drilling (2010)).
211
D PED Annex II diagrams
In the diagrams (1-9) of Annex II in PED, the maximum allow-able pressure (PS) (bar)
is plotted against the volume in liters, V(L) in the case of vessels and against the nominal
diameter (DN) in the case of pipes. These diagrams contain up to five subdivisions for the
various categories (sound engineering practice, I, II, III or IV). Delimiting curves on the
individual diagrams indicate the maximum values for the maximum allowable pressure
and the volume or nominal diameter in each category. These diagrams are presented in
Fig. D.1, D.2 and D.3.
214 D PED Annex II diagrams
1997L0023 — EN — 20.11.2003 — 001.001 — 30 1997L0023 — EN — 20.11.2003 — 001.001 — 31
▼B ▼B
Table 1 Table 2
Vessels referred to in Article 3, Section 1.1 (a), first indent Vessels referred to in Article 3, Section 1.1 (a), second indent
(a) Diagram 1: Vessels referred 1997L0023 to—in EN —Article
20.11.20033, (b)— 32Diagram 2: Vessels referred
— 001.001 1997L0023 to—in EN —Article
20.11.20033, — 001.001 — 33
▼B
Section
Exceptionally, 1.1(a), first
vessels intended indent.
to contain an unstable gas and falling within categories I or II on
▼B Section
Exceptionally, 1.1(a), second
portable extinguishers indent.
and bottles for breathing equipment must be classified at least
the basis of table 1 must be classified in category III. in category III.
Table 3 Table 4
Vessels referred to in Article 3, Section 1.1 (b), first indent Vessels referred to in Article 3, Section 1.1 (b), second indent
(c) Diagram 3: Vessels referred to in Article 3, (d) Diagram 4: Vessels referred to in Article 3,
Section 1.1(b), first indent. Section
Exceptionally, 1.1(b), second
assemblies intended indent.
for generating warm water as referred to in Article 3, Section
2.3, must be subject either to an EC design examination (Module B1) with respect to their confor-
mity with the essential requirements referred to in Sections 2.10, 2.11, 3.4, 5 (a) and 5 (d) of
Annex I, or to full quality assurance (Module H).
Figure D.1: Diagrams 1-4 of Annex II in PED
214
D PED Annex II diagrams 215
1997L0023 — EN — 20.11.2003 — 001.001 — 34 1997L0023 — EN — 20.11.2003 — 001.001 — 35
▼B ▼B
Table 5 Table 6
Pressure equipment referred to in Article 3, Section 1.2 Piping referred to in Article 3, Section 1.3 (a), first indent
(a) Diagram 5: Pressure equipment 1997L0023 — referred to in
EN — 20.11.2003 (b)— Diagram
— 001.001 36 6: Piping referred1997L0023 to in Article 3, Sec-— 001.001 — 37
— EN — 20.11.2003
▼B
Article 3,theSection
Exceptionally, 1.2. must be subject to a conformity assessment proce-
design of pressure-cookers
▼B
tion 1.3(a),
Exceptionally, pipingfirst
intendedindent.
for unstable gases and falling within categories I or II on the basis
dure equivalent to at least one of the category III modules. of Table 6 must be classified in category III.
Table 7 Table 8
Piping referred to in Article 3, Section 1.3 (a), second indent Piping referred to in Article 3, Section 1.3 (b), first indent
(c) Diagram 7: Piping referred to in Article 3, Sec- (d) Diagram 8: Piping referred to in Article 3, Sec-
tion 1.3(a),
Exceptionally, second
all piping containing indent.
fluids at a temperature greater than 350 ºC and falling within tion 1.3(b), first indent.
category II on the basis of Table 7 must be classified in category III.
Figure D.2: Diagrams 5-8 of Annex II in PED
215
216 D PED Annex II diagrams
1997L0023 — EN — 20.11.2003 — 001.001 — 38
▼B
Table 9
Piping referred to in Article 3, Section 1.3 (b), second indent
Figure D.3: Diagram 9 of Annex II in PED: Piping referred to in Article 3, Section 1.3(b),
second indent.
216
E PRV orifice sizes
The list of API and ASME orifice letters for SRV and corresponding effective area is given
in Table E.1. ASME orifice size denotes the actual orifice area whereas API the effective
orifice area.
Table E.1: Relief Valve Orifice Sizes

API Bore Dimensions ASME Bore Dimensions
Orifice letter
Orifice area [in ] Orifice area [cm ] Orifice area [in2 ] Orifice area [cm2 ]
2 2
D 0.110 0.71 0.1297 0.83

E 0.196 1.26 0.2279 1.47
F 0.307 1.98 0.3568 2.30
G 0.503 3.24 0.5849 3.77
H 0.785 5.06 0.9127 5.89
J 1.287 8.30 1.496 9.65
K 1.838 11.85 2.138 13.79
L 2.853 18.40 3.317 21.40
M 3.600 23.23 4.186 27.00
N 4.340 28.00 5.047 32.56
P 6.380 41.16 7.417 47.85
Q 11.05 71.29 12.85 82.90
R 16.00 103.22 82.60 120.00
T 26.00 167.74 28.62 184.64
F Vapor-liquid flow
Vapor-liquid flow is a commonly encountered situation is SRV, both in vertical (upstream)

and horizontal (downstream) direction. Therefore, although in this study only air-water
mixtures have been considered, a brief review of the physics and flow regimes in such
mixtures is given.
The formation of a two-phase mixture by vapor generation in a vertical heated tubular
channel represents an important special case. The presence of a heat flux through the
channel wall alters the flow pattern from the one which would have occurred in a long
unheated channel at the same local flow conditions. These changes occur due to two
main reasons; firstly, the departure from thermodynamic equilibrium coupled with the
presence of radial temperature profiles in the channel and secondly, the departure from
local hydrodynamic equilibrium throughout the channel. Figure F.1(a) shows a schematic
representation of a vertical tubular channel heated by a uniform low heat flux and fed at
its base with liquid just below the saturation temperature.
In the initial single-phase region the liquid is being heated to the saturation tempera-
ture. A thermal boundary layer forms at the wall and a radial temperature profile is set up.
At some position up the tube the wall temperature will exceed the saturation temperature
and the conditions for the formation of vapor (nucleation) at the wall are satisfied. Vapor
is formed at preferred positions or sites on the surface of the tube. Vapor bubbles grow
from these sites finally detaching to form a bubbly flow. With the production of more va-
por the bubble population increases with length and coalescence takes place to form slug
flow which in turn gives way to annular flow further along the channel. Close to this point
the formation of vapor at sites on the wall may cease and further vapor formation will
be as a result of evaporation at the liquid film-vapor core interface. Increasing velocities
in the vapor core will cause entrainment of liquid in the form of droplets. The depletion
of the liquid from the film by this entrainment and by evaporation finally causes the film
to dry out completely. Droplets continue to exist and are slowly evaporated until only
single-phase vapor is present (Collier and Thome (1994)).
Flow patterns formed during the generation of vapor in horizontal tubular channels
are influenced by departures from thermodynamic and hydrodynamic equilibrium in the
same way as for vertical flow. Asymmetric phase distributions and stratification introduce
additional complications. Fig. F.1(b) shows a schematic representation of a horizontal
tubular channel heated by a uniform low heat flux and fed with liquid just below the satu-
ration temperature. The sequence of flow patterns shown corresponds to a relatively low
inlet velocity (< 1 m/s). Important points to note from a heat transfer viewpoint are the
possibility of intermittent drying and rewetting of the upper surfaces of the tube in slug
and wavy flow and the progressive drying out over long tube lengths of the upper circum-
ference of the tube wall in annular flow. At higher inlet liquid velocities the influence
220 F Vapor-liquid flow
of gravity is less obvious, the phase distribution becomes more symmetrical and the flow
patterns become closer to those seen in vertical flow (Collier and Thome (1994)).
Figure F.1(b) illustrates the flow patterns existing during condensation inside horizon-
tal tubes. At the inlet film condensation around the circumference of the tube produces
an annular flow with some droplets entrained in the central high velocity vapor core. As
condensation continues, the vapor velocity falls and reduces the influence of vapor shear
on the condensate and the influence of gravity forces increases. At high flow rates slug
and bubble flows eventually are reached while at low flow rates large magnitude waves
and then stratified flow are formed. Reference studies in the topic are provided in Collier
(1972) and Collier and Thome (1994).
1)
ʓ߬˯ஏ
*%2 ,2
%͈
2)
( ᎟ଜഹᗕ ᎟ଝ߿ᗕ
2 ''2 + )-2
Ҽᗕ
'*#%2 '*#%)2
૗ᗕ Ì Íɲɲ ɲ
)2 +# *%2 '2
+# *%2 ,)2 ,2 ޼§§ॿ೗ *%2 '* 2
%-2 , 2 ()2*%2 ߷¥Łŏ೗ ,2 '*2 ,2 #*2 ,2 $*2
%͈
᎟ᗕଜ߿ᗕ ᎟ଜഹᗕ
3)
ʒࠓˮஏ
+# *%2 ' % 2 )% 2 %+)-2 )% 2
' %2 #% #2 %2

,2 ''2 + )-2 ,)2 ,+2 #% #) 2 %+ ) - 2 %2
'*#%/
)2
(*#% +# *%2
)2 ,),2 '
+ #!*%2 ')2 *%2
&.2 ,2 *&2 ߸¥Łŏ೗ ,2 ,+-2 ,2 ')%)2 ,2
ǁ͈
᎟ᗕଞᗕ߿ᗕ
re 7.10. The evolution(a) Steam/water flow(b)

of the steam/water in Steam/water flowtube.
a vertical boiler in a horizontal 1) evaporator tube, 2) condensation with high
flow in a vertical liquid loading and 3) condensation with low liquid loading
evaporator tube
Figure F.1: Vertical and horizontal evaporation-condensation in a tube (from Collier and
Thome (1994)).
re and contrast this flow pattern evolution with the inverted case of
tive boiling surrounding a heated rod in figure 6.4.
172
220
G Additional flow regime charts
The instances of vertical upward/downward cocurrent flow and slightly inclined pipe flow
are presented hereunder.
G.1 Vertical upward cocurrent flow

In this case, the axis is positively oriented in the upward direction and an upward cocurrent
flow occurs:
QG > 0 , QL > 0
G.1.1 Flow regimes

The main flow patterns that can be encountered during two-phase flow in a vertical upward
pipe, are presented in this paragraph. In Fig. G.1 we can see all these possible situations
of the flow. Hence, we can distinguish, by keeping constant the liquid flow rate and
increasing the gas flow rate, the following flow regimes:
• Bubbly flow
In this flow pattern, for low liquid velocity, the gas is dispersed in discrete bubbles.
These bubbles are entrained by the liquid. It is the most widely known configura-
tion, although it is not easily recognized for very high flow velocities.
• Disperse flow
As the liquid flow rate increases, the bubble diameter may increase via coalescence.
Bubbles are then crowded together and interact strongly with each other. The bub-
bly flow pattern is observed only in vertical and off-vertical flows in a relatively
large diameter pipes, whilst dispersed bubble flow is usually found over the whole
range of pipe inclinations. A difference between bubbly and disperse bubble flow
is found in distribution of bubble diameter since the first can be characterized as
polydisperse while the latter as monodisperse. However, distinction between the
two cases is not clearly visible.
• Slug flow
From bubbly flow, by a further increase of the gas flow rate, some of the bubbles
coalesce to form larger, long and cap-shaped bubbles. These large bubbles are also
termed Taylor bubbles. Therefore, slug flow is composed of a series of gas plugs or
222 G Additional flow regime charts
Taylor bubbles. The head of these plugs is generally blunt, whereas its end is flat
with a bubbly wake. Finally, by visual observation, we can notice that a thin liquid
film flows downwards, in respect to the pipe wall, around the Taylor bubbles.
• Churn flow
When increasing further the gas flow rate, another flow regime can be observed.
This is called churn (also called forth) flow. A lengthening and a breaking of the
gas plugs then occurs. This flow pattern evolves toward an annular flow in a chaotic
way. A difference between slug and churn flow is that the liquid falling film sur-
rounding the gas plugs does not appear anymore.
• Annular flow
Annular flow is characterized by the liquid flowing as a film around the pipe wall
and surrounding a high velocity gas core which may contain entrained liquid droplets.
Figure 7.6. flow
The upward A flow regime
of the liquidmap
film for the gravity
against flow of results
an air/water mixture
in from the forcesin a
exerted
vertical, 2.5cm diameter pipe showing the experimentally observed transi-
by the fast moving gas core. By visual observation, we can notice that droplets
tion
are regions
torn off hatched; the flow
from the crest of theregimes are propagate
waves that sketched inonfigure 7.7. Adapted
the surface of the liquid
from Weisman (1983).
film. They diffuse in the gas core and can eventually impinge onto the film surface.
According to Hewitt and Roberts (1969) wispy annular flow also can be formed in
the pipe, where the liquid droplets gather into clouds within the central gas core.
Figure G.1: Flow patterns in vertical upward flow in a pipe. Adapted from Weisman
(1983).
G.1.2 Flow maps

The flow regime map of Hewitt and Roberts (1969) is the most widely used chart for
upward oriented cocurrent air-water or steam-water flow in a pipe. This map is shown
in Fig. G.2. It was established initially for an air-water mixture flowing in a pipe of
31.2 mm diameter and at a static pressure range of 0.14 to 0.54 MPa. The Hewitt and
Roberts map was adapted several times in literature to fit with the flow conditions and
with experimental setup.
222
G.2 Vertical downward cocurrent flow 223
The coordinate system chosen from Hewitt and Roberts (1969) is:
Abscissa (Liquid kinetic energy):
2
G (1 − x)2
ρL U L2 = , (G.1)
ρL
Ordinate (Gas kinetic energy):
2
G x2
ρG UG2 = , (G.2)
ρG
where ρL and ρG are the density of the liquid and the gas respectively, U L and UG the
liquid and gas average velocities, G the mass velocity and x the quality. The abscissa of
this chart represents the liquid kinetic energy and the ordinate the gas kinetic energy. In
Fig. G.2, the flow pattern regions mainly addressed during this study (bubbly and bubble
slug flow) are highlighted.
In this point, we have to mention that the flow pattern maps are divided into two cate-
gories. The charts that are generated directly from experimental data, called experimental
flow maps and the mechanistic flow pattern maps which, in contrast, are developed from
the analysis of physical transition mechanisms and modeled by fundamental equations.
The latter ones are called theoretical flow maps. A typical example of experimental flow
map is the Hewitt-Roberts map (Fig. G.2) and an example of a theoretical flow chart is
the Taitel and Dukler (1977) map (Fig. G.5 in Appendix G.3).
Despite the numerous flow charts that exist in literature, we should use them with
caution because of the subjective definition of the flow regime by the different authors.
Therefore, the transition curves can differ a lot for each case. Finally, in order to take into
account the effect of the fluid properties and pipe diameter, additional correlations should
be introduced in the maps.
G.2 Vertical downward cocurrent flow

In this case, the axis is positively oriented in the downward direction. Thus, a downward
cocurrent flow will occur and we will have:
QG < 0, , QL < 0.
The vast majority of data collected for vertical two-phase flows has been confined to
upward cocurrent flow. However, in the recent years, the emergence of the deep-water
development has prompted designers to perform more intensive studies on downwards
cocurrent two-phase flow.
Since the gravitational and frictional terms in total pressure gradient calculation for
downward flow have opposite signs, gas-liquid downward flow in vertical pipes may ex-
perience either pressure loss or pressure gain, depending on flow rates, pipe geometry,
and fluid properties.
223
Gas kinetic energy ρGJG2 [kg/s2m]
Liquid kinetic energy ρLJL2 [kg/s2m]
Figure G.2: Vertical upward flow map of Hewitt and Roberts (1969).
G.2.1 Flow patterns

As previously mentioned, in the case of downward cocurrent flow, we can observe that the
separated flow region expands at the expense of the intermittent region. The intermittent
flow is restricted to a small region. As a result, the separated flow region has become
what is often called a “falling film” or “wetted wall” region. The distinction between this
region and the annular flow region is unclear (Cheremisinoff and Gupta (1983)).
An extensive review of all the possible flow patterns that can be met in downward
cocurrent flow can be found in Oshinowo and Charles (1974). Theses authors distin-
guished six different configurations. When comparing the flow regimes that are observed
in downward with the upward cocurrent two-phase flows, we can see that in downward
flow, we have, additionally, the falling film and the bubbly falling film regimes. Further-
more, the downward bubbly flow structure is quite different from the upward bubbly flow.
In the latter case, bubbles are spread throughout the whole tube uniformly, while for the
first case, the bubbles are oriented towards the center of the tube.
Finally, the annular flow regime can take several forms. For low liquid and gas flow
rates, a liquid film flows down the wall (falling film). When the liquid flow rate is in-
creased, the bubbly falling film configuration is observed, as the bubbles are entrained
within the film.
Example of flow regimes in downward cocurrent flow obtained in U-tube setup was
presented by Kourakos et al. (2007) and is demonstrated in Fig. G.3. Three visualization
points are also represented in flow map of Golan and Stenning (1969) in Fig. G.4; a
good agreement between experimental points and flow chart is remarked. Points 1 and 2
identified as bubbly dispersed and slug flow respectively belong to the region of slug and
224
G.3 Slightly inclined pipe flow 225
bubble flow of the chart while for point 3 flow pattern is classified as annular flow and
map predicts oscillatory flow (also not far from transition line to annular mist flow) for
the same conditions.
Bubble
W
QL=4.25 l/s,
JL=3.4 m/s,
GL=3390 kg/m^2s Volumetric quality:
1 7.3 % 2 10.2 % 3 25 %
A
QG=0.17 l/s, 0.5 %
JG=0.13 m/s,
GG=0.316 kg/m^2s
Bubble dispersed
W
QL=4.09 l/s,
JL=3.5 m/s
GL=3509 kg/m^2/s
A
QG=0.325 l/s
JG=0.26 m/s
GG=0.615 kg/m^2s
Slug
W
QL=1.45 l/s
JL=1.29 m/s
GL=1285 kg/m^2s
A
QG=0.17 l/s
JG=0.13 m/s
GG=0.32 kg/m^2s
Annular
W
QL=0.5 l/s
JL=0.53 m/s
GL=529 kg/m^2s Bubbly Bubbly dispersed Slug or churn Annular
A
QG=0.17 l/s
JG=0.13 m/s
Figure G.3: Flow regimes identified in vertical downward cocurrent flow taken from
GG=0.32 kg/m^2s
Kourakos et al. (2007).
G.2.2 Flow regime maps

An example of flow map for vertical downward cocurrent two-phase flow is that of Golan
and Stenning (1969). Comparing this map (Fig. G.4) to the Hewitt and Roberts (1969)
map (Fig. G.2), the annular flow regime seems to appear sooner considering the necessary
minimum gas velocity. Moreover, another region is introduced by Golan and Stenning
(1969), the oscillatory flow, which is an “intermediate” region. Another example of such
a map is the one proposed by Oshinowo and Charles (1974).
Taking an example of phase velocities U L =UG =1 m/s for both plots (vertical upward
and downward) the corresponding point is indicated as a big dot in charts of Hewitt and
Roberts and Golan and Stenning. It is obvious while for upward vertical flow bubbly and
bubbly slug are the dominant flow regimes, for downward flow the flow topology lies
between oscillatory and annular flow.
G.3 Slightly inclined pipe flow

Two-phase (gas-liquid) flow in downward inclined pipes is often stratified with the liquid
occupying the bottom part of the pipe. This flow regime is usually observed when gas and
liquid flow rates are sufficiently low.
In downward pipes at steep inclinations, Barnea et al. (1982) speculated that the mech-
anism by which stratified flow is transferred to annular flow at low gas flow rates is due
225
2.0
Annular Mist and

1.5 Annular Flow
Oscillatory
Flow
UG, m/s
1.0
Slug and
0.5 Bubble Flow
0.0
0.0 0.5 1.0 1.5 2.0
UL, m/s
Figure G.4: Vertical downward flow map of Golan and Stenning (1969).
to liquid droplets torn away from the wavy turbulent interface and thrown upon unwetted
tube walls, eventually leading to annular flow.
Taitel and Dukler (1976a) have derived a flow map for horizontal and slightly inclined
pipe from purely theoretical considerations. They distinguish between three classes of
flow: stratified flows in smooth or wave like form, intermittent flows in the form of slug
and plug flows and dispersed flows in the form of bubbly or annular-droplet flows. A con-
dition for the transition from stratified to intermittent flow is derived from the instability
condition for a soliton wave. The parameters involved in this map are non-dimensional
numbers that are defined as follows:
!1/2
ρG JG
F= , (G.3)
ρL − ρG (DgcosΘ)1/2
!1/2
ρG JG2 JL
K = F · ReL =
2
(G.4)
(ρL − ρG ) gνL cosΘ
and
!1/2
∇PL
T= (G.5)
(ρL − ρG ) gcosΘ
where νL the kinematic viscosity of liquid, X the Lockhart and Martinelli (1949) pa-
rameter (defined from Eq. 2.23 in §2.2.1.3), Θ the inclination angle of the pipe in respect
to the horizontal plane and with a positive orientation the vertical descending flow. ∇PL
is defined considering only the mass flow of the liquid according to the Lockhart and
Martinelli (1949) model and D the pipe diameter.
226
G.3 Slightly inclined pipe flow 227
8.2 Flow Models
For any model development it is very useful to divide two-phase flows ac-
cording to the scheme of Y. Taitel and A. Dukler (1976) into three classes:
separate flows, such as stratified flows, wavy flows and annular flows; inter-
mittent or transition flows in the form of elongated bubble flows, slug flows,
and plug flows; and dispersed flows like bubble flows, churn flows, and droplet
or mist flows. In order to describe two-phase flow, the mechanical coupling of
the state variables velocity, pressure, and temperature is usually carried out
in a Euler form of the conservation equations for mass, momentum, and en-
ergy. In the general case, the balance for each phase is taken separately, and
the description of the two-phase flow is then called a two-fluid model. This
procedure can generally also be applied to describe a flow with N fluids, and
it yields an N-fluid model. This has already been presented in general form
in Section 5.4.6. Next, we discuss the one-dimensional form of the two-fluid
model.
Fig. 8.4. Flow map in horizontal and slightly inclined pipe, after Y. Taitel and A.
Figure Dukler
G.5: Flow
(1976),map
withinthe
horizontal and numbers
characteristic slightly K,
inclined
F , andpipe
T asproposed
a functionby
of Taitel
the and
Martinelli parameter X
Dukler (1977) with the characteristic numbers K, F and T as a function of the Lockhart
and Martinelli (1949) parameter.
227
H Pressure drop in straight pipe SP
flow
The well known Darcy (1857) and Weisbach (1845) equation is a phenomenological for-
mula obtained by dimensional analysis to estimate the pressure drop along a straight pipe.
When the flow is established, the characteristics of the flow are independent of the
position along the pipe. The key quantities are then the pressure drop along the pipe per
unit length, ∆P/L, and the volumetric flow rate. The flow rate can be converted to an
average velocity U by dividing by the wetted area of the flow.
Pressure has dimensions of energy per unit volume. Therefore, the pressure drop
between two points must be proportional to (1/2)ρU 2 , which has the same dimensions
as it resembles (see below) the expression for the kinetic energy per unit volume. We
also know that pressure must be proportional to the length of the pipe between the two
points L as the pressure drop per unit length is a constant. To turn the relationship into
a proportionality coefficient of dimensionless quantity we can divide by the hydraulic
diameter of the pipe, D, which is also supposed to be constant along the pipe. Therefore,
L 1 2
∆P ∝ · ρU .
D 2
The so-called friction factor λ should be estimated and corresponds to proportional
coefficient of the previous formula.
Therefore, in single-phase flow, the friction term of the pressure drop of Eq. 2.22 is
given by the following equation:
L U2
∆P = λ ρ , (H.1)
D 2
where ∆P stands for the frictional pressure drop, λ is the Darcy friction factor, L the
length of the pipe, D the inner diameter of the pipe, ρ the density of the fluid and U the
mean velocity of the fluid. The latter is given by:
Q
U= , (H.2)
A
where Q is the flow rate and A the cross-sectional area of the duct.
As we can notice from Eq. H.1, in order to calculate the pressures drop, we have to
estimate the friction factor λ, since the rest of the terms of this equation are properties of
the fluid and parameters that can be measured (velocity and flow rate). There are many
correlations in literature for calculating the friction factor. The value of this parameter
230 H Pressure drop in straight pipe SP flow
depends on regime of the flow (laminar, transitional or turbulent) and on the roughness of
the surface of the pipe (smooth or rough).
Therefore, we must distinguish between the possible flow conditions, the laminar and
the turbulent flow:
Laminar flow: Re < 2300,
Turbulent flow: 4000 < Re < 105 .

It is obvious that the criterion for laminar or turbulent flow is the value of the Reynolds
number. We recall the definition of the Re number:
UD
Re = , (H.3)
ν
where ν the kinematic viscosity of the fluid.
Accordingly, for the case of laminar flow, we have:
64
λ= , (H.4)
Re
and for turbulent flow regime, the two correlations:
0.316
Blasius: λ= , (H.5)
Re0.25
!
1 2.51 k
Colebrook-White: √ = −2log √ + , (H.6)
λ Re λ 3.715D
where k stands for the roughness of the pipe.
The difference between the two formulas is that in the first case of the Blasius (1913)
law smooth pipe is considered whereas Colebrook and White (1937) accounts for rough-
ness of the pipe. We must call attention to the fact that the Blasius (1913) formula is
accurate up to about Re=105 . For a higher Re number the use of the Colebrook and White
(1937) formula is recommended; in the case of a smooth pipe a very low value of the duct
roughness should be used.
All the possible cases for the calculation of the friction factor λ are summarized in the
well known Moody’s diagram (Moody (1944)), which is shown in Fig. H.1. As we can
observe in Moody diagram, for important roughness of the pipe, friction factor is quickly
independent of Re number.
230
H Pressure drop in straight pipe SP flow 231
VD for water at 20°C (V in m/s, D in cm)

0.06 0.1 0.2 0.4 0.6 0.8 1 2 4 6 8 10 20 40 60 100 200 400 600 1000 2000 4000 6000 10000
|____________________________________________________________________________________________________________________________________________________________________________________________________
| | | | | | | | | | | | | | | | | | | | | | | | | | |
VD for atmospheric air at 20°C
1 2 4 6 8 10 20 40 60 100 200 400 600 1000 2000 4000 6000 10000 20000 40000 60000 100000
−1 | | | | | | | | | | | | | | | | | | | | | | | | | |
10
__
Laminar → Critical ← Transition zone →← Complete turbulence, rough pipes, R > 3500/r, 1/√f = 1.14 − 2 log r
9 flow zone
0.07
8
0.06
0.05
7
0.045
0.04
6 0.035
0.03
5.5
0.025
5
Relative roughness r Í Ä (ε in mm, D in mm)

0.02
Darcy−Weisbach friction factor f Í ÄÄÄÄ
2hDg
0.0175
2
4.5
LV
0.015
0.0125
4
0.01
3.5 0.008
0.006
3
D
ε
0.004
0.003
2.5
Material ε (mm)
ÄÄÄÄÄÄÄÄÄÄÄÄÄÄ ÄÄÄÄÄÄÄÄÄ 0.002
Riveted steel 0.9−9 0.0015
Concrete 0.3−3
2 Wood stave 0.18−0.9 0.001
Cast iron 0.25
0.0008
1.8 Galvanized iron 0.15
Asphalted cast iron 0.12 0.0006
Commercial steel 0.046 Smooth pipes, r __= 0
1.6 __ 0.0004
Drawn tubing 0.0015 1/√f = 2 log(R √f ) − 0.8
1.4 Fluid at 20°C ν (m2/s)
ÄÄÄÄÄÄÄÄÄÄÄÄÄÄ ÄÄÄÄÄÄÄÄÄ Hagen−Poisseuille equation 0.0002
Water 1.003e−006 R ≤ 2300, f = 64/R
Air (101.325 kPa) 1.511e−005
1.2 0.0001
Colebrook equation, R ≥ 2300 __
Latitude (WGS84) g (m/s2) __
ÄÄÄÄÄÄÄÄÄÄÄÄÄÄ ÄÄÄÄÄÄÄÄÄ 1/√f = −2 log(r /3.7 + 2.51/(R √f )) r = 5e−006

5e−005
10
−2 0.0° Sea level 9.78033
45.5° Standard 9.80665 Continuity equation, Q = AV
90.0° Sea level 9.83219 A = π D 2/4, V = 4Q /(π D 2) 2e−005
9
r = 1e−006
1e−005
8
6 7 8 3 2 3 4 5 6 7 8 4 2 3 4 5 6 7 8 5 2 3 4 5 6 7 8 6 2 3 4 5 6 7 8 7 2 3 4 5 6 7 8 8
10 10 10 10 10 10
VD
Moody Diagram Reynolds number R Í ÄÄ (V in m/s, D in m, ν in m 2/s)
ν
Figure H.1: Moody (1944) diagram.
231
I Straight pipe two-phase flow models
Additional two-phase flow (air-water and steam water) models for straight pipe are de-
scribed in this paragraph.
I.1 Steam-water Martinelli and Nelson (1948) model

An example of steam-water flow mixture method, proposed from Martinelli-Nelson, is
exemplified since although present investigation is focused in air-water flow, steam-water
is the common mixtures encountered during SRV operation.
The Lockhart and Martinelli (1949) method was extended for high pressures and tem-
peratures (i.e. steam-water mixtures) and was studied by Martinelli and Nelson (1948).
They have used the same formulas as the ones obtained by Lockhart-Martinelli for the
case of turbulent-turbulent regime.
The two-phase multiplier defined in this model is given by:
(∇P)GL
Φ2L0 =
(∇P)L0
where the subscript 0 defines that the conditions are calculated at the two-phase total
mass flux in contrast to the Lockhart-Martinelli model. In the basis of this definition, this
parameter is plotted against the mass quality for different pressures in Fig. I.1.
The relation between the two-phase multiplier of Lockhart-Martinelli and the multi-
plier defined by Martinelli and Nelson (1948) can be extracted:
Φ2L0 = (1 − x)1.75 Φ2L

When reaching the critical point of the vapor-liquid flow there is no distinction be-
tween the vapor and the liquid phase. Thus:
!1.75
1−x
µL = µG , ρL = ρG , Φ2L0 = 1, Xtt2 =
x
and
!1.75
1 + Xtt1.14
⇒ Φ2Ltt =
Xtt1.14
where tt denotes the turbulent-turbulent flow regimes for both air and water phase.
When studying two-phase flow with phase change, the contribution of the pressure
drop due to gravity and acceleration can be of significant importance and should be taken
234 I Straight pipe two-phase flow models
µ ò
ųʋʌƩӠ
$ ǄâŹǐ
ŴʡʢƪӠ
% & ' ( $

IY··Ó¦ÂYÁÓ ‫ ׻‬È ]ÌÓ³È
Figure I.1: Two-phase multiplier Φ2L0 for different mass qualities and pressures given by
Martinelli and Nelson (1948).
into account and added to the frictional one to obtain the total pressure drop. Hence, after
evaluating the contribution of each term Martinelli and Nelson have derived the following
equation for the total pressure drop:
G2 4L G2
∆P = (r4 ) L · g · cos (θ) + (r2 ) + (r3 ) fL0 (I.1)
ρL D 2ρL
where G the total (mixture) mass velocity, fL0 the friction factor of the liquid for
a mass flux of G, L the length of the pipe. The three factors r2 , r3 and r4 represent
the length-averaged two-phase pressure drop multipliers due to acceleration, friction and
gravity respectively. We should point out that the values of the three multipliers should be
evaluated for constant heating conditions and for the same models predicting Φ2L0 and α.
In Fig. I.2 the void fraction estimated with the model of Martinelli-Nelson for different
mass qualities is shown.
For both the HEM and Martinelli-Nelson models the following assumptions are made
by Todreas and Kazami (1989):
• The influence of the mass flux is negligible for a constant quality; this result is
not verified experimentally since the two-phase flow multipliers are found to be
dependent of the mass flow rate.
• The surface tension effect is not taken into account in the model; this is not accurate
for high pressure conditions.
234
I.2 Drift-flux model 235
QXyr«¶ÈÈZ¶È ³ È
4 ް৊ࢤੂ઎োࢥம 4 4

4 -4
˜ÔиӠ ůПёχÔƦӠ
.4
"4 #4
ƀīƈ‫ڌ‬ம
‫׻׌‬
࡯ம & '4
4
Ӿ‫׻‬
4
4 4
o
×ò 4
ĺ‫׻‬
4
4
Ë؆‫ڃ‬ம
/4 4
!4
!4 4
ǆம ǆம
4 04 4
IX¦¦ÈXyr«¶È ‫׻‬
Figure I.2: Void fraction versus mass quality for different pressures by Martinelli and
Nelson (1948).
I.2 Drift-flux model

A detailed description of the drift-flux model can be found in Zuber and Findlay (1965)
and Wallis (1969). In this approach, the vapor velocity is expressed as the sum of the local
two-phase volumetric velocity J and the local drift velocity of the vapor Uv j (Eq. I.2).
Jv = αUv = αJ + (Uv − J) . (I.2)

When averaging this summation, we obtain:
hJv i2 = hαJi2 + hα (Uv − J)i2 (I.3)

| {z }
Drift-flux term
where notation hi2 indicates area averaging. The drift-flux term is indicated in this
formula it represents the rate at which the vapor passes through a unit area that is already
traveling with the total flow at velocity J.
The ratio of void fraction α to volumetric quality β can be written with the drift flux
(Todreas and Kazami (1989)):
hαi2 1
= (I.4)
hβi2 C0 + Uv j
hJi2
with
hαJi2
C0 = .
hαi2 hJi2
and
235
Uv j = hα (Uv − J)i2 / hαi2

where C0 the concentration parameter that represents the effect due to non-uniform
void fraction and velocity profiles. Uv j is the local relative velocity effect.
Correlation I.4 can be written in the form:
hαi2 = K hβi2 (I.5)

The values of K and C0 obtained for different models and authors are summarized in
Table I.1
Table I.1: Values of coefficient K and C0 for drift flux model proposed by several authors.
Model-author Drift-flux terms K, C0 Limitations-conditions
HEM C0 =1, Uv j '0 High flow rates, low
slip ratio
Armand and Treschev K=0.833+0.005ln (10P), P in Low slip ratio
(1959) Pa, Uv j '0
Bankoff (1960) K=0.71+0.001P where P in psi, Low slip ratio
Uv j '0 β
Dix (1971) C0 = hβi2 1 + hβi − 1 , β =
1
All flow regimes, low
2
ρ 0.1
v
slip ratio
ρl , Uv j '0
If Eq.I.4 is transformed in terms of mass flow rate, mass quality and is combined with
the definition of C0 , the slip ratio can be expressed as:
(C0 − 1) xρl Uv j ρl
S = C0 + + (I.6)
(1 − x) ρv (1 − x) G
where the second term on the right-hand side of this equation represents the variation
of the slip ratio due to non-uniform void fraction distribution and the third term is due to
the local difference in the velocity of vapor and liquid.
When a uniform void distribution takes place C0 =1. Values of C0 close to zero corre-
spond to low void fraction while when this parameter approaches unity, the void fraction
is relatively high. Zuber and Findlay (1965) proposed C0 =1.2 for bubbly and slug flow.
Ishii (1977) have expanded the drift-flux model for annular flow.
I.3 Steam-water Thom (1964) model

The work performed by Thom (1964) was concentrated on the estimation of the same
coefficients r2 , r3 and r4 of the Martinelli and Nelson (1948) model with additional data
mainly for high pressures. The results obtained and the values of the aforementioned
parameters proposed by Thom (1964) are illustrated in Fig. I.3.
As a concluding remark, one can state that the Martinelli and Nelson (1948) and
Thom (1964) correlations give reliable findings for a mass velocity up to ' 1000 kg/m2 s
while the HEM model works better for the highest mass fluxes of ' 2000 kg/m2 s and
236
I.4 Baroczy model 237
r4
r2
r3
Figure I.3: Acceleration multiplier r2 (left), gravitational multiplier r4 (right up) and fric-
tional multiplier r3 (right down) function of the operating pressure for different mass
qualities given by Thom (1964).
higher. This can be explained by the fact that for a low mass flow rate and a given quality
an annular pattern (separated flow) occurs and therefore the first two models are more
precise (steam-water mixtures). However, for a higher mass flux the two phases are more
dispersed (bubble or slug flow) and in these conditions the HEM model is more trustful.
I.4 Baroczy model

Baroczy (1966) has adopted a model to correct the influence of total mass flux on the
two-phase multiplier. His method can be used for water but also mercury, sodium and
Freon-22. The correlation, established by Baroczy (1966), considered fluid properties,
mixture mass quality and mass flux. The liquid to gas viscosity and density ratio was
237
called as the property index. Therefore, as shown in Fig. I.4, φ2L0 is plotted in function of
the property index (for a mass velocity G0 =1356 kg/m2 s):
!0.2
µL ρG
!
.
µG ρL
+(.$. ੑঘৈ‫ڙ‬ம ܿம ੐ம
˵ǩɻʞʱ±Ӡ џʹ¡ΰйӠΔўӠʓƥӠ
iʩ
˨‫׻‬
ІĘ‫׻‬
Ϩ
ҽӠ .
ͭӠ
ƞǐ $
_v
I‫׻‬
ϱϲ‫׻‬ ܾம Ϗ‫׻‬
֤‫׻‬ ‫׾‬ம
ȡӠ
ٖம .
ଆம
‫ݙ‬
$ iʩ
«Ӡ
ֺĪ‫ض‬ம ॓‫ࢅͫޞ‬ம ɸͩ‫؀‬ம ɼaǪʕӠ «aǫɽӠ
Nb¤¶È uabµÈ ÃćvŎFe&/ÈvŎĽvŎPeÈ
",. ‫܇‬৉ࢢচॿͬŤŤம
$ $
#)((,. .
̲ÿ͵ψ΃ωѷϼӠ Ű̦ˣƧӠ
%+&-. . $ $ $
*'. .
$ $ $
$
$
Figure I.4: Baroczy (1966) two-phase flow multiplier.
The advantage of the definition of this parameter was that it did not require knowledge
of the critical pressure and temperature in order to establish the property ratios at the
critical point, where they had a value of 1.
Another curve of the correcting factor of φ2L0 was also plotted (Fig. I.5) when the mass
flux in the channel is not the reference mass flux G0 . This correcting plot is a function of
the physical property index, quality and mass velocity.
I.5 Chisholm (1976) model

Chisholm (1976) has based his studies on the Lockhart and Martinelli (1949) model and
derived the following equation:
n o
Φ2L0 = 1 + Γ2 − 1 B [x (1 − x)](2−n)/2 + x2−n (I.7)
238
I.6 Evaluation of the models 239
ƞமǻʩ ±ʷʾӠ . ‫ؘ‬Դٕ‫ځ‬ம ऴࣦமɭܽம
ࡴம YǗம
গம
ƛǐ cம
ĊӠ
ՆՇம
Íò cȩ×ம
͔‫׻‬
ׄ‫׻‬
ҫ‫׻‬
՗‫׻‬ ƞɑமŬŬ‫ډ‬ம !.
.
˂˃‫׻‬
҈҉‫׻‬
ٔம
Ɯǐ
.ȧ‫ؗ‬ம
ųՂம հӼ‫׻̪ ׻‬
֓c͵ųம
Figure I.5: Baroczy (1966) curves for mass flux correction factor.
+2
2−n
where Γ , (dP/dz) G0
(dP/dz)L0
, B , CΓ−2
Γ2 −1
, C the Lockhart and Martinelli (1949) parameter
and n=0.25 (Blasius (1913) constant). This formula has been proposed as a sufficiently
accurate method for engineering purposes.
I.6 Evaluation of the models

Todreas and Kazami (1989) compared the different models of two-phase flow and have
plotted the results in Fig. I.6. Three different values of the mass flux were investigated;
G1 =4000 kg/m2 s, G2 =400 kg/m2 s and G3 =40 kg/m2 s.
The conclusions extracted from this analysis is that the homogeneous flow model is
mainly accurate for high pressure and flow rates values. The Martinelli and Nelson (1948)
model is good in the range of 500 < G (kg/m2 s) < 1000. Finally, when the drift-flux model
for slug flow is used, it is concluded that the model is closer to the correct assumed flow
regime (bubbly and churn patterns) for the lowest mass flux G3 .
The study of Idsinga et al. (1977) was focused on the comparison of different two-
phase pressure drop multiplier models. In total, eighteen methods were tested for 2200
adiabatic steam-water points and 1230 under diabatic conditions. For a big variety of flow
239
These results are plotted in Figure 1 1 - 1 3 .
To determine how appropriate these correlations are , w e need to consider the
flow conditions and the flow regimes associated with G " G 2 , and G3 •
1 .0 Homogeneous for G1 , G2 and G3
0.9
]:
C
.2
� 0.8
u.
"
0.7
0.6 ;------+--��--+_--�-
0.5 0.6 0.7 0 .8 0.9 1 .0
Quality, x
Figure 11·13 Variation of the void fraction with quality of Example 1 1 -3.
Figure I.6: Comparison of different two-phase flow models (Todreas and Kazami (1989)).
conditions and flow patterns, the authors concluded that the most appropriate models have
been proved the HEM, Baroczy (1966) and Thom (1964) models.
Table I.2 shows a review of the models that were evaluated as the most accurate from
several authors (EDSU (1976), Friedel (1977)).
Table I.2: Comparison of two-phase pressure drop correlations for steam-water mixtures
by Idsinga et al. (1977).
5ʾdþԷ [1976]
B]]lŎ
þѽ ͊Է ƗFō (Է[1977]
ilŎ úqԷ
ĒĢ ƇԷ Է LUԷ [1977]
B>1jŎ ¦Ü Է ¦Ü Է
cƨ qԷ ǷԷ þǷ `Է ōƨ Է ?ƨ Է
DqԷ \‫׻‬ Ã‫׻‬ ‫׻‬ \‫׻‬ Ҙ‫׻‬ ‫׻‬ \‫׻‬ ҙ‫׻‬ ‫׻‬ ×‫׻‬ Ã‫׻‬ ‫׻‬
l5ĉԷ ;ம !<Ŏ uGV Ŏ 1 ; Ŏ

ú5DԷ Ŏ F;GŎ uF! Ŏ ! Ŏ ;6Ŏ Û FG G Ŏ . Ŏ !<Ŏ u1V Ŏ < Ŏ .!F?Ŏ Ü? Ŏ !.L 1 Ŏ
\ƨ`Է[1966]
i.lŎ F..;Ŏ . Ŏ ! 6Ŏ ;6Ŏ ÝFëK 1Ŏ ?1# ; Ŏ !<Ŏ u< < Ŏ G# ; Ŏ .?!Ŏ ÞK Ŏ ? c < Ŏ
[1976]
GōFqƠsʾ6q Է B1lŎ F6!1Ŏ FG Ŏ ?1 Ŏ ì;6Ŏ ß? < Ŏ ?1c Ŏ ?<Ŏ 6 Ŏ . # 6 Ŏ .!F!Ŏ ?æGŎ !. V . Ŏ
ĉ UsfF Է [1948]
;jŎ >.Ŏ F1ç?Ŏ ?1c 1 Ŏ
!<Ŏ .;K < Ŏ .!K ;Ŏ
=9ĢYŎ KԷ 08ƠԷ GqqōȁԷ[1972]

i<ĕŎ
Gɑqō FŇԷ ×‫ ׻ ׻‬6ƠǵԷɃԷԷ ō FԷ `ƨϴԷ YŎ ‫ ׻‬Ơ Է 8ˮԷ ̠ǑԷ eʩ ʘӳò࡮ࡊ॑ம ͤ ӷӸò࢟સদ˕ம ú‫׻‬
£vŎӴòࢠહধ‫ڭ‬ம ‫ ׻ ׻‬F Է7ſſ Է 0ԷɑFԷ &6Է Է Ơ Է Ǒ A Է
I.7 Flow pattern dependent models

The pressure drop correlations classified by different flow patterns were recapitulated and
evaluated by Beattie (1973). Table I.3 summarizes these results.
240
I.7 Flow pattern dependent models 241
Table I.3: Comparison of flow regime dependent models by Beattie (1973) (taken frrom
Delhaye et al. (1980)).
λ
In this table, friction factor λ and the Re number are associated with formula (Cole-
brook and White (1937)):
!
1 2 9.35
√ = 3.48 − 4log + √ (I.8)
λ D Re λ
where the roughness of the pipe √ and Re number is defined including the quality X.
For the wall-bubble model Re=2.4 We with We the Weber number defined as the ratio
of the inertia to surface tension.
An extensive review of the pressure drop and void fraction correlations for various
flow patterns can be found in Govier and Aziz (1972).
241
J Single-phase pressure change in
singularities
Example of pressure drop coefficients for various singularities from Idel’Cik (1986) is
demonstrated in Fig. J.1.
Smooth expansion
Sudden expansion
Smooth contraction
Sudden contraction
Figure J.1: Different types of singularities and the proposed pressure drop coefficients by
(Idel’Cik (1986)).
An additional summarizing table of the range of the pressure loss coefficient for vari-
ous pipe fittings, bends and valves, taken from Todreas and Kazami (1989) is exemplified
in Table J.1
244 J Single-phase pressure change in singularities
J.1 Abrupt area changes

Explanation of the pressure change occurrence in abrupt area changes in incompressible
fluid is shown in Fig. J.2.
Calculation of pressure change in sudden expansions and contractions for incompress-
ible and compressible fluids is exemplified in Table J.2.
J.2 Calculations in geometries with change of section

When the diameter of the pipe is changing, a change in the velocity of the fluid arises.
This leads to a difference in the dynamic pressure as well, in the two cross-sections. The
latter has to be taken into account in the determination of the pressure drop of the system.
A simplified sketch of a smooth contraction is given in Fig. J.3.
As it is known, the total pressure is given by the summation of three components, the
Table J.1: Pressure loss coefficient range for different singularities (taken from Todreas
and Kazami (1989))
gaÿ ň
rĩþň ഒ Ry 9ň
r ň nň un &ň ň y2&ň

./ / ň ? *!!ň >ň ň ň (ԥ(ഒ 7 ň ň
; / X4 x ň Q!!ň 9ň ň ň Zഒ ň ň
;vP Â!!ň a n9ň (2ഒ 7ň ň
C? 9$ň $ ň 6>ň xǕഒ 7ň ň
CM ň J M ň 6 ň ň N*&ň
3ň ň Jň X (ഒ 7ň ň
;2!!ň 4ň ň9 ¤ /ň nň

; *!!ň 9 6 ô9© ň x X 2 ഒ͐ ċ ഒƟ ȄࡔͰ̎‫̏޵ڐ‬ഒ à Ĵ&ň
; Q!!ň J ň ̿ ĉ ഒƠ Ȅ ̌ ҥ ࡔ í ¡&ň
&YYň 9 ĸÃ96 Yň ?ň
4aň ԑ࠷
NPň S6M /ň ?ň
gą?9ň
rP&Pň ̼ഒࡼഒ‫ص‬ഒ߶ഒͱৡରઆઇ੅॒ഒ ഒ R/ ň
_ °ň
þ౓ഒ ;!!ň / L ň Zഒ Z2U(þഒ âň ň
þ౔ഒ è Ä?!M * { ň NL ň ഒ (ഒ Иx_ഒ 7ň ė ň
2˭ഒ ;!!ň / L ň X_ഒ (y ƕ Uöഒ ãň aň
; !?!ň ň | đ ň 4? *4ň * ~ ň ഒ (ۭഒ U_ഒ 7 ň m ň
;!,!ň Łň | Ē ň 4, 24ň L 9 v p ň _ഒ x_öU ƕ ഒ äň ň
TxRň } R 2ň pň

* ğ ň ň X öU X 2 Õഒ åň ň
y 9N !ň Ɩ Z ഒ Й2ഒ 7ň ň
҈ӛ 3 ? J$ &ň R/*×ň *Ġ6ň 7!ġ4M5ň े Ɩ Z ‫ڑ‬ഒ k ň J$Raň L* Y 6 $ Kň 3Y Æň ň T 7 7 ň $Rň ň
Idel'Cik [1986]
**Values
2 ň u ň K depend
of4ň on the
4 ¡/ň Lpipe diameter
ň u ň L ň Ԥ ഒ
**n=1 approximate for ň
Re>10 4
± T/ *ň j ňഒ ! !ň 64ň ; $n=0.75 more exact.
ň ! m &eň
¯ň ࠷ Ǽഒ J &ň - ň gň Ԛ࠷
$ ň Sӛ Õ2ഒ & ň Jqň
244
J.2 Calculations in geometries with change of section 245
</?1/959?
ʣ‫ڦ‬ Ǘ‫ڦ‬ 31$/9?

Г
5)5:?9)1/?
>1/?
7359$1/?31%/9?
89!//9? *:$?
@̦ @̦
ֵ‫ڦ‬
- - °
99&? 577:4? 4?

!ā

‫ڦ‬

/ Ӿ‫ڦ‬ ѻ ‫ڦ‬
/

7+13? ÎĤ ɒ ֶɓ
Ĥ

®ą \· ? ͌‫ڦ‬
9? 1 ? </?
1/949?
,bH·1 ·1 .1a+·
··1/1.1/1.1/1 1.1.1. /1/·10·-1·
!!Ĥ
Figure J.2: Demonstration of the pressure changes in incompressible fluid at abrupt con-
traction and expansion geometries (taken from Todreas and Kazami (1989)).
(1) (2)
d1 u1 u2 d2
∆PM
Figure J.3: Simplified schematic of pressure drop in LUCY II.
static, the dynamic and the hydrostatic pressure:
Ptot = P st + Pdyn + Phydrostatic . (J.1)
Therefore, we have:
1
Ptot = P st + ρU 2 + ρgh, (J.2)
2
where P st the static pressure, ρ the density, U the velocity and h the height of the fluid.
245
246 J Single-phase pressure change in singularities
Table J.2: Pressure changes in abrupt contraction and expansion geometries in incom-
pressible and compressible fluid (taken from Todreas and Kazami (1989)).
,ƻ =` .ƻ ^ .ƻ
^ ƻ ƻ Ù. éƻ

9 ƻ ?L Ĥ ƻ ? L Ĥ
%³ʒʽ %³ʓʽ
l
H6012L Ĥ ƻ E C032L Ĥ ƻ
?@A+L Ĥ ƻ ?BC,/JL Ĥ ƻ
Q ƻ -Ѝ ‫ڦ‬ Q ƻ -ę ‫ڦ‬

ƻ ?!!L Ĥ ƻ ?L Ĥ %% % %
΋ַÆ֨։֊‫ڦ‬ ƻ '<7;L ƻ
]^2·
Ĥ Ĥ
£‫ ڦ‬% əÐǰíʽ ¸Ĥ
ď̨ȗƴ‫ڦ‬
ÔĀþÿĤ r
n
Ĥ
D+" L %
Ğӆ‫ڦ) ڦ‬ ǸȒ‫) ڦ‬Ğө‫ڦ‬
Q ƻ ħį-ʐЎÓ‫ڦ‬ Q ƻ ħį-‫ڦ‬ę ‫ڦ‬
M3ƻ ƻ ƻ ƻ ª

ƻ ƻ §ĳƻ
d.` ƻ
?# L
õĤ
o %
· %
8ʽ Ģ 8ʽ
· ?!/JL % %
·ą M°·
!%
p

?E=89L Ĥ ƻ ?FD451L Ĥ ï‫ڦ‬

?BC,$L Ĥ ƻ (C-!L Ĥ ƻ
QEƻ -) ‫ڦ‬ Q ƻ ¢-ļÓ‫ڦ‬

`ƻ
?L
öĤ
q %
#% $
Ųʽ 8ʽ
·
K?
?B:%L
L Ĥ
Ĥ ƻ
Q ƻ Ěʑͥ ‫ڦ‬
%
"%
»Ĥ )‫ڦ‬
?GC+!L Ĥ
Q ƻ KĚKļ ‫ڦ‬

ï‫ڦ‬
FĤ
ďƆÆρϑ‫ ̦ ڦ¸¸ؠץ‬ƻ aƻ ŤŻ ƻ ƻ ƻ Ï0ƻ
)*I
.#JL Ĥ ƻ Ũֈ‫` ڦ‬ƻ ªƻ ÕPƻ 2Ź
ƻ ťƌƻ
ƻ ƻ ;ţÔŔƻ ;ƻ Õªźƻ Ũ؊‫ ڦ‬.ƵÓƻ ƻ Ԁԁ
ƆÆ֩‫פ‬գբ¸٧‫ڦ‬
J.2.1 Single-phase pressure drop

As we can see from Fig. J.3, the term of the hydrostatic pressure could be neglected
for our case, as we are measuring the average over the pipe in four points with a 90 ◦
angle between them. The static pressure is given by the measurement of the pressure
transducers. Finally, the dynamic pressure could be calculated if we know the velocity of
the fluid in the pipe.
An average velocity before and after the singularity could be assumed (Flow rate
given by the flow meters). This could be calculated by Eq. 2.17. This quantity is called
superficial velocity. For a cylindrical pipe, the following formula could be used:
4Q
U= . (J.3)
πd2
Therefore, for the case of single-phase flow, the pressure drop is (Eq.J.2):
1 2
∆P(12)
tot,S P = P(2)
st − P (1)
st + ρ U 2 − U 2
1 . (J.4)
2
And after replacing all the terms:
246
J.2 Calculations in geometries with change of section 247
8ρ d14 − d24
" #
∆P(12) = ∆P M + 2 10−6 Q2 , (J.5)
tot,S P
π (d1 d2 )4
where ∆P M the static pressure drop in Pa, ρ the density of the fluid in kg/m3 , d1 the
diameter of the pipe upstream and d2 its diameter downstream of the restriction in m and
Q the volumetric flow rate of the fluid in l/s. It is obvious that the only two variables of
this equation are; ∆P M and Q.
J.2.2 Two-phase pressure drop

The calculation of the two-phase pressure drop is performed in the same way with the
difference that an average density and flow rate of the two-phase mixture should be cal-
culated (Eq. 2.18).
Using an average superficial velocity J ∗ (according to Eq. 2.17), volumetric flow rate
of the mixture Q∗ and by adapting Eq.J.5 for the case of two-phase flow, we conclude to
the following formula that is giving the final total pressure drop:
8 d14 − d24
" #
∆Ptot,T P = ∆P M + 2
(12)
10−6 ρ∗ (Q∗ )2 . (J.6)
π (d1 d2 ) 4
In this case, the variables are: ∆P M , ρ∗ and Q∗ .
J.2.3 Pressure drop coefficient in contractions

For the pressure drop coefficient in contractions, Comolet (1963) has proposed the fol-
lowing simple formulas for smooth and sudden contraction respectively:
!2
1 π
ζ= − 1 sinα for α < (J.7)
CC 2
!2
1 π
ζ= −1 for α > (J.8)
CC 2
where ζ is the total pressure drop coefficient, CC the contraction coefficient given by
Eq. 2.36 and α the angle of the smooth singularity.
247
K Pressure drop database
In this paragraph, summarizing tables of all pressure drop coefficient results obtained in
singularities and SRV are presented.
K.1 Expansion singularities

A summarizing table with two input two output parameters is built; upstream total Reynolds
number Ret1 , volumetric quality β and total singular pressure drop ∆Ptot
sing
and dimension-
less pressure drop coefficient ζ respectively. The total Re number is calculated with the
formula:
Jt1 · d
Ret1 = (K.1)
νL
where νL the liquid viscosity, d the upstream diameter and Ut1 the total upstream
velocity ((QL +QG )/A1 ).
Table K.1 refers to abrupt area expansion of σ=0.43.
Table K.1: Results in sudden expansion σ=0.43.
Sudden enlargement, d=0.041 mm, D=0.0627 mm, σ=0.43

sing
Input Ret1 · 104 [-] β [%] Output ∆Ptot [mbar] ζ [-]
8.37 0 -6.1 0.344

18.20 0 -25.1 0.288
21.90 17 -33.8 0.389
23.40 0 -41.1 0.287
24.00 4 -41.7 0.296
8.29 0 -5.9 0.326
8.68 5 -6.4 0.356
8.98 7 -6.9 0.381
9.38 11 -13.9 0.765
9.87 0 -8.2 0.318
10.30 4 -8.6 0.336
10.70 8 -9.4 0.365
11.20 11 -11.3 0.437
11.70 15 -19.2 0.739
12.10 0 -11.5 0.297
Continued on Next Page. . .
250 K Pressure drop database
Table K.1: (continued)
sing
12.60 4 -12.4 0.319

13.00 7 -13.2 0.341
13.70 11 -14.4 0.372
14.10 14 -27.7 0.724
16.60 0 -18.7 0.262
17.10 3 -19.9 0.278
17.80 7 -21.2 0.296
18.30 10 -22.5 0.315
19.00 13 -24.1 0.338
19.70 16 -25.7 0.360
20.50 19 -28.3 0.394
21.60 24 -38.8 0.540
20.00 0 -29.3 0.283
20.50 3 -30.6 0.295
21.10 5 -31.1 0.301
21.80 9 -33.1 0.321
22.50 12 -34.4 0.333
23.30 15 -35.9 0.348
24.20 17 -38.3 0.367
24.50 19 -38.7 0.372
24.10 0 -41.4 0.291
24.70 3 -42.3 0.297
25.40 5 -43.2 0.305
26.10 8 -45.2 0.318
26.80 11 -46.7 0.329
27.60 13 -48.4 0.341
28.20 15 -49.2 0.347
In table K.2 the results obtained for sudden expansion of σ=0.65 are reported. Com-
pared to the previously demonstrated configuration of smaller upstream diameter and sur-
face area ratio, considerably lower values of pressure drop coefficient are detected for
identical Re number. This can be explained by the lower dynamic pressure difference.
At this point it should be pointed in all presented results, ζ is calculated with subtrac-
tion of regular loss. Therefore, the physical meaning of this parameter is found in pressure
drop caused only due to presence of singularity.
Table K.2: Results in sudden expansion σ=0.65.
Sudden enlargement, d=0.0627 mm, D=0.078 mm, σ=0.65

sing
13.40 0 -2.7 0.140

250
K.1 Expansion singularities 251
sing
18.10 0 -4.2 0.116

20.10 6 -3.9 0.103
23.20 0 -6.3 0.114
24.70 6 -5.6 0.101
16.70 0 -2.6 0.092
17.10 2 -2.3 0.083
17.50 5 -3.0 0.108
18.30 0 -2.8 0.083
18.70 2 -2.9 0.086
19.20 4 -2.8 0.083
19.60 7 -6.8 0.202
19.40 0 -3.4 0.088
19.80 2 -3.6 0.094
20.20 4 -3.3 0.087
20.70 6 -4.2 0.110
21.20 8 -15.2 0.398
20.40 0 -5.0 0.119
20.80 2 -4.8 0.115
21.30 4 -4.9 0.118
21.70 6 -5.1 0.121
22.30 8 -12.0 0.285
23.20 0 -7.3 0.131
23.70 2 -7.2 0.129
24.10 3 -7.0 0.126
24.70 5 -6.8 0.122
25.20 7 -7.0 0.126
25.80 9 -7.7 0.138
Review of all results acquired in smooth expansion α=5 ◦ is given in Table K.3.
Table K.3: Results in smooth expansion σ=0.43, α=5 ◦ .
Smooth enlargement, d=0.041 mm, D=0.0627 mm, σ=0.43, α=5 ◦

sing
7.98 0 -1.9 0.102

18.20 0 -8.3 0.094
20.10 8 -14.1 0.160
23.10 0 -12.0 0.084
25.00 8 -14.1 0.099
10.20 0 -2.9 0.114
9.83 2 -4.3 0.169
251
sing
9.97 4 -5.9 0.233

10.20 5 -6.8 0.264
10.40 7 -8.3 0.322
11.80 0 -4.1 0.107
11.70 2 -5.1 0.133
12.00 3 -7.2 0.188
12.30 5 -8.8 0.227
12.60 7 -10.1 0.262
13.00 9 -11.4 0.296
16.40 0 -7.0 0.097
17.10 2 -7.1 0.099
17.60 3 -7.3 0.100
17.80 5 -8.6 0.118
17.70 7 -10.3 0.142
18.00 9 -13.7 0.189
18.60 11 -16.8 0.231
19.80 0 -9.2 0.089
20.30 2 -7.7 0.075
20.80 4 -9.8 0.095
21.20 6 -10.5 0.101
21.80 8 -13.0 0.126
22.20 10 -16.2 0.157
22.80 12 -19.7 0.190
23.50 0 -9.8 0.068
24.10 3 -10.5 0.073
24.80 5 -12.3 0.086
25.60 8 -13.7 0.095
26.30 10 -16.7 0.116
27.10 13 -21.6 0.150
27.50 14 -23.9 0.166
Table K.4 outlines the extracted pressure findings for similar flow conditions in smooth
expansion of opening angle α=8 ◦ .
Smooth enlargement, d=0.041 mm, D=0.0627 mm, σ=0.43, α=8.2 ◦

sing
7.72 0 -2.7 0.147

17.50 0 -11.7 0.132
19.20 8 -16.7 0.188
252
sing
23.70 0 -17.7 0.123

25.70 8 -19.9 0.138
9.57 0 -3.9 0.148
9.71 1 -5.2 0.200
10.10 3 -6.2 0.235
10.20 5 -6.7 0.257
10.50 7 -7.5 0.288
11.90 0 -5.2 0.136
12.20 2 -6.7 0.174
12.40 3 -7.8 0.204
12.70 5 -8.6 0.222
12.80 7 -9.9 0.257
12.80 9 -11.2 0.290
15.90 0 -9.4 0.130
16.30 2 -9.5 0.132
16.80 4 -10.6 0.147
17.30 5 -11.8 0.163
17.70 7 -14.2 0.197
18.10 9 -16.2 0.226
18.70 12 -18.0 0.251
19.20 0 -12.8 0.123
19.60 2 -12.0 0.116
20.40 4 -13.5 0.130
20.70 6 -14.5 0.141
21.40 8 -17.4 0.168
22.50 10 -19.9 0.192
23.10 12 -23.3 0.225
24.10 0 -17.4 0.121
24.50 2 -15.5 0.109
24.00 4 -17.3 0.120
24.60 6 -18.4 0.128
25.20 8 -20.0 0.139
25.90 10 -22.8 0.158
26.40 11 -25.8 0.179
Finally, in Table K.5 a synopsis of imposed flow conditions and measured experimen-
tal and calculated values for smooth expansion geometry of α=15 ◦ is pointed up.
Smooth enlargement, d=0.041 mm, D=0.0627 mm, σ=0.43, α=14.8 ◦

sing
253
sing
8.04 0 -4.2 0.231

17.50 0 -19.9 0.225
19.40 8 -21.6 0.243
23.60 0 -33.1 0.231
25.30 8 -35.5 0.247
9.87 0 -5.8 0.223
10.60 4 -6.4 0.247
11.20 9 -8.5 0.325
12.40 0 -8.7 0.225
13.20 5 -9.7 0.251
14.10 10 -11.8 0.302
17.60 0 -16.9 0.235
18.70 5 -17.9 0.248
19.70 9 -19.9 0.275
20.70 15 -24.3 0.337
21.90 20 -28.8 0.397
19.10 0 -24.0 0.232
20.30 5 -25.6 0.246
22.00 10 -27.6 0.265
23.40 15 -32.2 0.310
24.80 17 -35.9 0.345
23.70 0 -34.1 0.238
24.80 5 -35.1 0.245
26.20 10 -36.9 0.258
27.50 14 -40.8 0.285
Deviations from literature-expansion singularities

In Table K.6 deviations of pressure drop coefficient ζ obtained from Idel’Cik (1986) for
single-phase flow for all expansion geometries, Chisholm (1976) and Jannsen and Kervi-
nen (1966) for two-phase sudden expansion geometries are presented in detail.
Table K.6: Deviation experimental-literature correlation for ex-

pansion.
Conditions Deviation in %
Geometry ReL1 · 104 [-] β [-] Idel’cik Jannsen Chisholm
[1986] [1966] [1969]
σ=0.43, α=90 ◦ 8.37 SP -4.84 - -

18.20 SP 13.60 - -
254

[1986] [1966] [1969]
23.40 SP 14.20 - -
8.29 SP 0.62 - -
9.87 SP 3.12 - -
12.10 SP 10.20 - -
16.60 SP 25.10 - -
20.00 SP 15.60 - -
24.10 SP 12.70 - -
24.10 2.52 - 13.30 17.60
20.00 2.70 - 14.10 18.70
16.60 3.23 - 21.90 27.80
12.10 3.58 - 6.61 12.30
23.20 3.73 - 15.10 21.50
9.86 3.88 - 1.49 7.36
8.28 4.59 - -3.61 2.99
24.00 5.26 - 13.50 22.40
20.00 5.36 - 15.20 24.40
16.60 6.67 - 18.70 30.50
12.10 6.92 - 3.32 14.00
8.31 7.40 - -7.05 3.21
9.90 7.59 - -2.90 8.10
24.00 8.02 - 12.10 25.50
19.90 8.84 - 12.10 26.90
16.50 9.90 - 15.40 32.40
24.00 10.50 - 11.40 28.90
12.20 10.90 - -1.22 14.80
8.34 11.10 - -51.80 -43.80
9.92 11.30 - -15.50 -1.28
19.90 11.60 - 11.40 30.60
24.00 13.10 - 10.60 32.20
16.50 13.20 - 11.90 33.90
12.10 14.00 - -47.40 -36.30
19.90 14.50 - 10.30 34.10
9.94 14.90 - -47.80 -36.20
24.00 14.90 - 11.10 35.80
16.50 16.20 - 8.66 35.00
18.20 16.90 - 1.35 26.90
20.00 17.40 - 7.98 35.90
20.00 18.50 - 8.15 38.00
16.60 19.20 - 2.93 32.40
16.50 23.50 - -20.70 7.19
σ=0.65, α=90 ◦ 13.40 SP -10.30 - -
18.10 SP 7.76 - -
23.20 SP 9.65 - -
255

[1986] [1966] [1969]
16.70 SP 35.50 - -
18.30 SP 51.60 - -
19.40 SP 43.10 - -
20.40 SP 4.90 - -
23.20 SP -4.65 - -
23.30 1.64 - -1.06 4.59
20.40 1.92 - 11.30 19.30
19.40 2.01 - 36.00 45.80
18.30 2.13 - 49.10 61.00
16.70 2.35 - 54.90 68.20
23.30 3.29 - 2.55 14.70
20.40 3.88 - 10.80 26.70
19.40 4.03 - 49.90 72.50
18.30 4.46 - 58.10 83.20
16.70 4.66 - 21.60 42.30
23.40 5.05 - 7.78 27.20
19.00 5.69 - 28.40 55.00
20.40 5.92 - 9.82 33.70
19.40 6.18 - 21.20 48.80
23.10 6.42 - 32.50 63.50
18.30 6.68 - -33.70 -17.70
23.40 6.94 - 7.22 34.40
20.50 7.85 - -52.20 -38.60
19.50 8.24 - -65.70 -55.40
23.40 9.03 - -0.57 32.30
σ=0.43, α=5 ◦ 7.98 SP -0.72 - -
18.20 SP 1.84 - -
23.10 SP 11.80 - -
10.20 SP -13.10 - -
11.80 SP -8.36 - -
16.40 SP -1.24 - -
19.80 SP 6.37 - -
23.50 SP 37.70 - -
σ=0.43, α=8.2 ◦ 7.72 SP -4.99 - -
17.50 SP 2.93 - -
23.70 SP 9.82 - -
9.57 SP -6.68 - -
11.90 SP 1.42 - -
15.90 SP 4.61 - -
19.20 SP 10.70 - -
24.10 SP 11.10 - -
σ=0.43, α=14.8 ◦ 8.04 SP 6.59 - -
256
K.2 Contraction singularities 257

[1986] [1966] [1969]
17.50 SP 8.81 - -
23.60 SP 5.46 - -
9.87 SP 10.00 - -
12.40 SP 8.79 - -
17.60 SP 3.96 - -
19.10 SP 5.28 - -
23.70 SP 2.58 - -
K.2 Contraction singularities

Pressure drop coefficient ζ determined at sudden contraction of σ=2.34 is reported for all
tested conditions in Table K.7.
Table K.7: Results in sudden contraction of σ=2.34.
Sudden convergence, d=0.0627 mm, D=0.041 mm, σ=2.34

sing
18.50 0 -44.0 0.222

19.20 5 -52.0 0.263
22.00 0 -60.9 0.219
24.20 9 -78.0 0.281
16.40 0 -32.6 0.208
16.90 3 -35.5 0.227
17.20 5 -38.1 0.244
17.60 7 -43.4 0.277
18.10 10 -52.5 0.336
18.80 13 -61.9 0.395
19.20 15 -58.0 0.371
18.00 0 -38.3 0.206
18.50 3 -41.7 0.224
18.90 5 -46.5 0.251
19.30 7 -49.7 0.266
19.90 10 -61.9 0.331
20.60 13 -70.3 0.376
20.90 14 -67.3 0.362
19.10 0 -43.1 0.204
19.70 3 -46.8 0.221
20.10 5 -51.1 0.242
20.60 7 -56.8 0.268
21.20 10 -66.8 0.316
257
sing
21.80 11 -77.0 0.364

22.00 12 -76.0 0.358
20.10 0 -46.9 0.201
20.70 3 -52.5 0.224
21.20 5 -56.6 0.241
21.60 7 -61.0 0.260
22.30 10 -71.2 0.303
22.70 11 -77.2 0.329
21.90 0 -56.5 0.203
22.60 3 -61.1 0.219
23.10 5 -66.1 0.237
23.60 7 -71.8 0.258
24.30 9 -79.0 0.283
Considerably smaller values of pressure drop coefficient are extracted for smooth ge-
ometry of σ=2.34 and angle α=15 ◦ . The exact acquired values of ζ and total pressure
drop are reviewed in Table K.8 for different Ret1 and volumetric qualities of air.
Table K.8: Results in smooth contraction of σ=2.34, α=15 ◦ .
Sudden convergence, d=0.0627 mm, D=0.041 mm, σ=2.34, α=14.8 ◦

sing
18.70 0 -5.1 0.026

19.70 5 -8.2 0.041
22.40 0 -5.2 0.019
24.70 9 -11.0 0.040
16.60 0 -2.0 0.013
17.20 3 -2.2 0.014
17.60 5 -2.6 0.016
17.90 7 -2.9 0.018
18.60 10 -4.0 0.026
19.20 13 -4.0 0.026
19.70 15 -1.8 0.011
18.20 0 -1.7 0.009
18.80 3 -2.1 0.011
19.20 5 -2.9 0.016
19.60 7 -3.2 0.017
20.20 10 -3.7 0.020
20.90 13 -3.6 0.020
21.30 14 -2.8 0.015
19.50 0 -1.0 0.005
258
K.2 Contraction singularities 259
sing
20.10 3 -3.0 0.014

20.50 5 -3.1 0.015
21.00 7 -3.5 0.017
21.70 10 -4.0 0.019
22.40 12 -3.4 0.016
22.40 13 -3.9 0.018
20.50 0 -0.5 0.002
21.10 3 -2.3 0.010
21.70 5 -3.3 0.014
22.10 7 -3.9 0.017
22.80 10 -4.4 0.019
23.20 11 -4.5 0.019
22.50 0 -0.4 0.002
23.20 3 -2.8 0.010
23.80 5 -3.7 0.013
24.20 7 -4.8 0.017
25.00 10 -5.6 0.020
Finally, smooth contraction of different surface area ratio σ=1.56 and α=9 ◦ has been
investigated. Pressure drop coefficient for this case is summarized in Table K.9
Table K.9: Results in smooth contraction of σ=1.56, α=9 ◦ .
Smooth convergence, d=0.040 mm, D=0.032 mm, σ=1.56, α=9.1 ◦

sing
5.56 0 -5.1 0.271

7.13 0 -6.1 0.198
8.91 0 -6.9 0.144
10.70 0 -8.5 0.122
12.50 0 -10.9 0.116
14.30 0 -13.2 0.107
15.30 0 -14.0 0.098
9.36 14 -13.2 0.274
9.29 12 -13.3 0.275
9.20 9 -12.7 0.263
9.14 7 -11.9 0.248
9.06 5 -11.3 0.235
9.00 3 -10.8 0.225
8.96 2 -10.3 0.214
8.93 1 -9.8 0.203
12.90 10 -17.0 0.180
12.90 9 -17.0 0.180
259
sing
12.80 7 -16.5 0.175

12.70 5 -16.6 0.176
12.60 4 -15.5 0.164
12.60 2 -15.0 0.158
12.50 1 -14.4 0.152
12.50 0 -14.4 0.152
15.80 8 -20.4 0.143
15.70 7 -19.1 0.134
15.60 5 -18.4 0.129
15.60 4 -17.8 0.125
15.50 3 -17.6 0.123
15.40 2 -17.0 0.119
15.40 1 -15.9 0.111
15.30 0 -15.0 0.106
Deviations from literature-contraction singularities

In Table K.10 deviations of experimental results from calculated literature coefficients are
reported.
Table K.10: Deviation experimental-literature correlation for con-

traction.
Conditions Deviation in %
Geometry ReL1 · 104 [-] β [-] Idel’cik [1986] Jannsen [1966]
σ=2.34, α=90 ◦ 18.50 SP 28.80 -

22.00 SP 30.60 -
16.40 SP 37.30 -
18.00 SP 39.20 -
19.10 SP 40.20 -
20.10 SP 42.70 -
21.90 SP 41.30 -
16.40 2.75 - 21.40
20.20 2.84 - 23.00
21.90 2.84 - 26.00
18.00 2.85 - 23.30
19.10 2.97 - 24.70
16.40 4.57 - 15.10
18.30 4.72 - 7.05
22.00 4.88 - 19.10
20.20 4.90 - 16.80
19.10 4.92 - 16.50
17.90 4.99 - 12.60
260
K.3 SRV pressure drop database 261
Geometry ReL1 · 104 [-] β [-] Idel’cik [1986] Jannsen [1966]
18.00 6.70 - 8.02

20.20 6.72 - 10.40
16.40 6.77 - 3.61
22.00 6.80 - 11.40
19.20 7.01 - 7.61
22.00 8.82 - 4.61
22.00 9.49 - 4.66
20.20 9.56 - -2.05
18.00 9.57 - -10.60
19.10 9.58 - -6.25
16.40 9.62 - -11.70
20.20 11.10 - -8.41
19.20 12.20 - -16.00
19.20 12.40 - -14.40
18.00 12.50 - -18.60
16.40 12.80 - -22.20
18.00 14.10 - -13.70
16.40 14.80 - -15.20
σ=1.56, α=9.1 ◦ 8.91 SP -55.10 -
10.70 SP -48.10 -
12.50 SP -45.80 -
14.30 SP -41.90 -
15.30 SP -37.10 -
K.3 SRV pressure drop database

Pressure drop coefficient results are reported in Table K.11
Table K.11: Results in transparent SRV at P set =0.15 MPa. .
Transparent SRV at P set =0.15 MPa

Input L [mm] Ret1 · 104 [-] β [%] Output ζ [-]
22.49 0 1.84
22.08 2 1.93
21.23 5 2.00
20.14 10 2.12
7.3
19.37 14 2.22
18.39 20 2.37
16.59 33 2.76
13.68 52 4.03
21.10 0 2.04
20.95 2 2.10
6.5
261
20.21 5 2.17
19.20 10 2.32
18.54 14 2.44
18.17 20 2.57
20.27 0 2.33
20.14 2 2.38
19.30 5 2.51
5.5
18.28 10 2.70
17.69 14 2.83
16.90 20 3.02
19.44 0 2.59
19.91 2 2.61
18.66 5 2.90
4.5
17.61 10 3.19
17.00 14 3.36
16.22 20 3.61
19.38 0 2.64
18.40 2 2.82
17.26 5 3.31
4
15.94 10 3.69
15.87 14 3.99
14.75 20 4.28
18.49 0 2.98
18.16 2 3.16
16.77 5 3.68
3.5
15.66 10 4.05
15.14 14 4.38
14.42 20 4.69
17.10 0 3.62
16.15 2 4.30
15.29 5 4.71
3
14.45 10 5.08
13.96 14 5.40
13.23 20 5.98
16.95 0 4.20
14.86 2 4.92
14.12 5 5.34
2.6
13.83 10 5.80
13.36 14 6.15
12.50 20 7.04
12.73 0 7.17
12.20 2 8.12
11.56 5 8.84
2
10.97 10 9.62
262
K.3 SRV pressure drop database 263
10.54 14 10.34
10.44 20 10.61
9.98 0 9.78
9.69 2 11.64
9.34 5 12.28
1.5
9.01 10 13.23
8.66 14 14.17
11.15 20 15.41
7.57 0 29.30
7.37 2 21.16
7.06 5 21.85
1.0
6.85 10 23.70
6.67 14 24.92
6.59 20 26.59
263
L Matlab codes
The MatLab ® codes used for processing of experimental results obtained in geometrical
singularities and safety relief valve tested are illustrated hereunder. Additionally, in the
codes, computations with the theoretical models analyzed and derived in Chapter 2 are
presented whilst input and output calculated parameters are indicated.
L.1 Singularities
Input:
d [m], D [m], length of singularity l s [m], Angle of singularity [ ◦ ], ∆P sing st [mbar], t
[ C], Ql [l/s], P1 [mbar], β [%], mg [g/s], Reattachment/vena contracta length [-].
◦
Output:
d [m], D [m], Angle [ ◦ ], Gt1 [kg/m2 s], Jt1 [m/s], ζ [-], ζliterature [-], Deviation-Idel’cik
[%], Deviation-Jannsen [%], Deviation-Chisholm [%], ∆Ptot sing [mbar], Ret1 [-], ReL1 [-],
Rem1 [-], Rem2 [-], β [%], L/d [-], ρ [kg/m ], ρL [kg/m ], ρG [kg/m3 ], x [-], µm [Pa·s], ul
∗ 3 3
[m/s].
Code:
1 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
2 %%%%% Program to process singularities %%%%%
3 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
4 close all;
5 clear all;
6 %--------------------------------------------------
7 %% input data and geometry file
8 root='C:\Data\PhD\RESULTS\Singularities_data';
9 %%%% Load xls!!!!!!!
10 % ext='.xls'
11 % for i=1:42
12 % filename=[root,ext]
13 % fid=fopen(filename,'r');
14 % file=xlsread(filename, 1, 'A3:K44');
15 %%% Load txt!!!!!!!
16 ext='.txt';
17 for i=1:305
18 filename=[root,ext];
19 fid=fopen(filename,'r');
20 file=dlmread(filename,'',[1 0 305 10]);
21 fclose(fid);
266 L Matlab codes
22 d=file(i,1);
23 D=file(i,2);
24 ls=file(i,3);
25 alpha=file(i,4);
26 DPst_sing=file(i,5);
27 t=file(i,6);
28 q_l=file(i,7);
29 P1=file(i,8);
30 beta=file(i,9);
31 m_g=file(i,10);
32 reat_length=file(i,11);
33 %%% Subscript 1 is upstream, 2 is downstream, g is gas,
34 %%% l is liquid and t is total (gas and liquid),
35 %%% st is static and total is dyn and static.
36 %%% d is upstream diameter, D is downstream diameter
37 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
38 %%% To be loaded from file: d [m], D [m], ls [m],
39 %%% alpha, DPst_sing [mbar],
40 %%% t [C], q_l [l/s], P1 [mbar], beta [%], m_g [g/s],
41 %%% Reat_ven_contr length
42 %%% L/d [-]
43 %% Loop
44 if alpha==90
45 angle=90;
46 else
47 angle=atan((abs((d-D))/2)/ls)*(180/pi);
48 end
49 A1=pi*(d^2)/4;
50 A2=pi*(D^2)/4;
51 sigma=A1/A2;
52 T=t+273;
53 %% Properties
54 dyn_visc_water=0.00000000011252*(t^4) - 0.000000018249*(t^3)
55 + 0.0000013193*(t^2) - 0.000059044*(t) + 0.0017836;
56 v_air_at_20C=1.51E-05;
57 rho_l=-0.000000416943*(t^4) + 0.0000766621*(t^3)
58 - 0.00905352*(t^2) + 0.0763421*(t) + 999.841;
59 v_water=dyn_visc_water/rho_l;
60 rho_g=(P1*100)/(287.05*T);
61 %%Mixture properties
62 dyn_visc_air=v_air_at_20C*rho_g;
63 k_visc=dyn_visc_air/dyn_visc_water;
64 rho_cor=(beta/100)*rho_g+(1-(beta/100))*rho_l;
65 %% Parameters
66 q_g=m_g/rho_g;
67 q_t=q_l+q_g;
68 m_l=(q_l/1000)*rho_l;
69 m_t=m_l+(m_g/1000);
70 u1_g=(q_g/A1)/1000;
71 u1_l=(q_l/A1)/1000;
72 u1_t=(q_t/A1)/1000;
73 u2_g=(q_g/A2)/1000;
74 u2_l=(q_l/A2)/1000;
75 u2_t=(q_t/A2)/1000;
76 Gt1=m_t/A1;
266
L.1 Singularities 267
77 Gt2=m_t/A2;
78 Re1_l=(u1_l*d)/v_water;
79 Re2_l=(u2_l*D)/v_water;
80 Re1_g=(u1_g*d)/v_air_at_20C;
81 x=(m_g/1000)/m_t;
82 %Mixture viscosity Einstein
83 mu_m=dyn_visc_water*(1+2.5*(beta/100)
84 *((k_visc+0.4)/(k_visc+1)));
85 %%%%%%%%%
86 Re1_m_2=(rho_cor*u1_t*d)/dyn_visc_water;
88 Re1_m=(rho_cor*u1_t*d)/mu_m;
90 Re1_t=(u1_t*d)/v_water;
91 Re2_t=(u2_t*D)/v_water;
92 %% Loop-expansion-contraction
93 if d<D
94 DPtot_sing=((100*DPst_sing)-(0.5*rho_cor
95 *((u1_t^2)-(u2_t^2))))/100;
96 Zeta=-DPtot_sing/((0.5*rho_l*(u1_l^2))/100);
97 else
98 DPtot_sing=-((100*DPst_sing)+(0.5*rho_cor
99 *((u1_t^2)-(u2_t^2))))/100;
100 Zeta=-DPtot_sing/((0.5*rho_l*(u2_l^2))/100);
101 end
102 %% Test correlation
103 % x=Zeta;
104 % y=beta;
105 % mx=mean(x);
106 % my=mean(y);
107 % mxy=mean(x.*y);
108 % % Standard Dev. from built-in Matlab Functions
109 % std(x,1)
110 % std(y,1)
111 % % Standard Dev. from Equation Above
112 % sqrt( 1/5 * sum((x-mx).^2))
113 % sqrt( 1/5 * sum((y-my).^2))
114 % COVARIANCE=cov(x,y,1);
115 % CORRELATION=corrcoef(x,y);
116 %% Literature correlations %%%%%%%
117 %%% Lambda calculation
118 if Re1_t<10^4 % Blasius
119 lambda1=0.3164/(Re1_t^0.25);
120 lambda2=0.3164/(Re2_t^0.25);
121 else
122 %Absolute roughness
123 k=1.50E-06;
124 r1=k/d;
125 r2=k/D;
126 lambda1=0.1*(((1.46*r1)+100/Re1_t)^0.25);
127 lambda2=0.1*(((1.46*r2)+100/Re2_t)^0.25);
128 end
129 %%% Regular loss
130 %%%Define:L1 length L/d upstream, L2 length L/d downstream
131 L1=8;
267
268 L Matlab codes
132 L2=reat_length;
133 DP_reg_1=(lambda1*L1*0.5*rho_cor*u1_t^2)/100;
134 DP_reg_2=(lambda2*L2*0.5*rho_cor*u2_t^2)/100;
135 %% Expansion
136 if d<D
137 %%% Idelcik
138 if beta==0
139 %%% Sudden enlargement
140 if angle==90
141 zeta_lit=(1-sigma)^2;
142 deviation0=((zeta_lit-Zeta)/Zeta)*100;
143 deviation1=0;
144 deviation2=0;
145 %%% Smooth enlargement
146 else
147 zeta_f=lambda1/((8*sin((pi/180)*(angle)))*(1-(sigma)^2));
148 zeta_el=(3.2*(tan((pi/180)*(angle))))
149 *((sqrt(tan((pi/180)*(angle))))^(1/4))*((1-sigma)^2);
150 zeta_lit=zeta_el+zeta_f;
152 deviation1=0;
153 deviation2=0;
154 end
155 %%% Two-phase correlations
156 else
157 %%% Sudden enlargement
158 if angle==90
159 %%% Jannsen
160 DPtot_jannsen=-(((Gt1^2)/(2*rho_l))*((1-sigma)^2)
161 *(1+x*((rho_l/rho_g)-1)))/100;
162 zeta_lit1=-DPtot_jannsen/((0.5*rho_l*(u1_l^2))/100);
163 %%% Chisholm
164 X=sqrt(((1-x)/x)^2*(rho_l/rho_g));
165 C=(1+0.5*((rho_l-rho_g)/rho_l)^(0.5))*((rho_l/rho_g)^(0.5)
166 +(rho_g/rho_l)^(0.5));
167 DPst_chisholm=(((Gt1^2)/rho_l)*sigma*(1-sigma)*((1-x)^2)
168 *(1+(C/X)+(1/(X^2))));
169 DPtot_chisholm=((DPst_chisholm)-(0.5*rho_cor*
170 ((u1_t^2)-(u2_t^2))))/100;
171 zeta_lit2=-DPtot_chisholm/((0.5*rho_l*(u1_l^2))/100);
172 deviation1=((zeta_lit1-Zeta)/Zeta)*100;
173 deviation2=((zeta_lit2-Zeta)/Zeta)*100;
174 deviation0=0;
175 if abs(deviation1)<abs(deviation2)
176 zeta_lit=zeta_lit1;
177 else
178 zeta_lit=zeta_lit2;
179 end
180 %%% Smooth enlargement
181 else
182 %%%% Find correlation for smooth expansion
183 zeta_jannsen=0;
184 zeta_chisholm=0;
185 zeta_lit=0;
186 deviation0=0;
268
L.1 Singularities 269
187 deviation1=0;
188 deviation2=0;
189 end
190 end
191 %% Contraction
192 else
193 %%% Idelcik
194 if beta==0
195 %%% Sudden contraction 65/40
196 if angle==90
197 zeta_lit=0.5*(1-(1/sigma));
199 deviation1=0;
200 deviation2=0;
201 %%% Smooth contraction 15 degrees
202 elseif 14<angle<16
203 zeta_prime=0.1031*((ls/D)^(-0.3951));
204 zeta_f=lambda2/((8*sin((pi/180)*(angle)))*(1-(1/sigma)^2));
205 zeta_lit=zeta_prime*(1-(1/sigma))+zeta_f;
207 deviation1=0;
208 deviation2=0;
209 %%% Smooth contraction 9 degrees 40/32
210 elseif 8<angle<10
211 zeta_prime=0.15631*((ls/D)^(-0.3026));
212 zeta_f=lambda2/((8*sin((pi/180)*(angle)))*(1-(1/sigma)^2));
213 zeta_lit=zeta_prime*(1-(1/sigma))+zeta_f;
215 deviation1=0;
216 deviation2=0;
217 end
218 %%% Two-phase correlations
219 else
220 %%% Sudden contraction
221 if angle==90
222 %%% Weisbach contraction coefficient C_c
223 C_c=0.63+0.37*((1/sigma)^3);
224 DPtot_jannsen=-(((Gt2^2)/(2*rho_l))
225 *(((1/C_c)-1)^2)*(1+x*((rho_l/rho_g)-1)))/100;
226 zeta_lit=-DPtot_jannsen/((0.5*rho_l*(u2_l^2))/100);
228 deviation0=0;
229 deviation2=0;
230 %%% No correlation for Chisholm for sudden contraction
232 else
233 %%%% Find correlation for smooth contraction
234 zeta_jannsen=0;
236 zeta_lit=0;
237 deviation0=0;
238 deviation1=0;
239 deviation2=0;
240 end
241 end
269
270 L Matlab codes
242 end
243 %% Results and plots
244 %%%Initinalize with zeros
245 %%% REMARK: deviation0--> Results-Idelcik,
246 %%% deviation1--> Results-Jannsen,
247 %%% deviation2--> Results-Chisholm
248 Results(i,:)=[d, D, angle, Gt1, u1_t, Zeta, zeta_lit,
249 deviation0, deviation1, deviation2, DPtot_sing, Re1_t,
250 Re1_l, Re1_m, Re1_m_2, beta, reat_length, rho_cor,
251 rho_l, rho_g, x, mu_m, u1_l];
252 end
253 close all;
254 hdr={'d [m]', 'D [m]', 'Angle [^\circ]', 'Gt1 [kg/m^2s]',
255 'Jt1 [m/s]', 'Zeta [-]', 'Zeta literature [-]',
256 'Deviation-Idelcik [%]', 'Deviation-Jannsen [%]',
257 'Deviation-Chisholm [%]', 'DPtot_sing [mbar]', 'Re1_t [-]',
258 'Re1_l [-]', 'Re1_m [-]', 'Re1_m_2 [-]', 'Beta [%]',
259 'L/d [-]', 'rho_cor [kg/m^3]', 'rho_L [kg/m^3]',
260 'rho_g [kg/m^3]', 'x mass_qual', 'mu_m [Pas]', 'ul [m/s]'};
261 txt=sprintf('%s\t',hdr{:});
262 txt(end)='';
263 dlmwrite('Results.dat',txt,'');
264 dlmwrite('Results.dat', Results,'-append','delimiter',
265 '\t','precision',3);
266 % xlswrite('Results',txt,'');
267 % xlswrite('Results',Results);
L.2 SRV
Input:
d [m], D [mm], P1,rel [Bar], L [mm], Temperature [ ◦ C], Force [N], Qwater [l/s], Qair
[g/s], P1,abs [Bar], P2,abs [Bar], β [%] Condition (flashing or non-flashing), Condition
(transparent-industrial SRV).
Output:
P set [bar], Mass quality x [-], F [N], L [mm], G1,exp [kg/m2 s], Gω [kg/m2 s], GHNEDS
[kg/m2 s], GAPI [kg/m2 s], Transparent/Metallic SRV, ρ∗ [kg/m3 ], Critical/subcritical con-
dition.
Code:
1 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
2 %%%%% Program to process metallic/transparent SRV %%%%%%%%%%%%
3 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
4 close all;
5 clear all;
6 %------------------------------------------------------
7 %% input data SRV
8 root='C:\Data\PhD\Safety_valve\All_Processing
9 \SRV_data_flashing';
10 % root='C:\Data\PhD\Safety_valve\All_Processing
270
L.2 SRV 271
11 \SRV_data_non_flashing';
12 ext='.txt';
13 % for i=1:170
14 for i=1:279
17 % file=dlmread(filename,'',[1 0 170 12]);
19 fclose(fid);
20 d=file(i,1);
21 D=file(i,2);
22 P1_rel=file(i,3);
23 L=file(i,4);
24 t=file(i,5);
25 F=file(i,6);
26 q_l=file(i,7);
27 m_g=file(i,8);
28 P1=file(i,9);
29 P2=file(i,10);
30 beta=file(i,11);
31 cond=file(i,12);
32 cond_transp=file(i,13);
33 %%% Subscript 1 is upstream, 2 is downstream,
34 %%% g is gas, l is liquid and t is total (gas
35 %%% and liquid), st is static and total is dyn
36 %%% and static. d is upstream diameter, D is
37 %%% downstream diameter, uploaded pressures
38 %%% are total pressures
39 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
40 %%% To be loaded from file: d [m], D [mm], P1 rel [Bar],
41 %%% L [mm], Temp [C], Force [N], Water [l/s], Air [g/s],
42 %%% Pabs1 [Bar], Pabs2 [Bar], Beta [%], Condition [-],
43 %%% Condition_transp [-]. ALL units are in SI!!!
44 %% Properties-parameters-Experimental results
45 A1=pi*(d^2)/4;
46 A2=pi*(D^2)/4;
47 sigma=A1/A2;
48 T=t+273;
49 %% Properties
50 dyn_visc_water=0.00000000011252*(t^4) - 0.000000018249*(t^3)
51 + 0.0000013193*(t^2) - 0.000059044*(t) + 0.0017836;
52 v_air_at_20C=1.51E-05;
53 rho_l=-0.000000416943*(t^4) + 0.0000766621*(t^3)
54 - 0.00905352*(t^2) + 0.0763421*(t) + 999.841;
55 v_water=dyn_visc_water/rho_l;
56 rho_g=(P1*100000)/(287.05*T);
57 %%Mixture properties
58 dyn_visc_air=v_air_at_20C*rho_g;
59 k_visc=dyn_visc_air/dyn_visc_water;
60 rho_cor=(beta/100)*rho_g+(1-(beta/100))*rho_l;
61 %% Parameters
62 q_g=m_g/rho_g;
63 q_t=q_l+q_g;
64 m_l=(q_l/1000)*rho_l;
65 m_t=m_l+(m_g/1000);
271
272 L Matlab codes
66 u1_g=(q_g/A1)/1000;
67 u1_l=(q_l/A1)/1000;
68 u1_t=(q_t/A1)/1000;
69 u2_g=(q_g/A2)/1000;
70 u2_l=(q_l/A2)/1000;
71 u2_t=(q_t/A2)/1000;
72 Gt1=m_t/A1;
73 Gt2=m_t/A2;
74 Re1_l=(u1_l*d)/v_water;
75 Re2_l=(u2_l*D)/v_water;
76 Re1_g=(u1_g*d)/v_air_at_20C;
77 x=(m_g/1000)/m_t;
78 %Mixture viscosity Einstein
79 mu_m=dyn_visc_water*(1+2.5*(beta/100)*((k_visc+0.4)
80 /(k_visc+1)));
81 %%%%%%%%%
86 Re1_t=(u1_t*d)/v_water;
87 Re2_t=(u2_t*D)/v_water;
88 %% Literature correlations in two-phase flow for SRV %%%
89 % % Omega method
90 %%% All pressures transformed in absolute pressures!!!
91 k=1.001;
92 N2=1;
93 N3=1;
94 v_g1=1/rho_g;
95 v_l1=1/rho_l;
96 v_1=1/rho_cor;
97 v_gl1=v_g1-v_l1;
98 Cp=4184;
99 hgl1=2200000;
100 n2=P2/P1;
101 %% %%%%%% Omega method %%%%%%%%%%%
102 %% Determine omega
103 %%% Non-Flashing
104 %%% 1=Non-Flashing
105 if cond==1
106 omega=(x*v_g1)/(v_1*k);
107 %%% Flashing
108 %%% 0=Flashing
109 elseif cond==0
110 omega=((x*v_g1)/(v_1*k))+((N2*Cp*T*(P1*100000))
111 /v_1)*((v_gl1/hgl1)^2);
112 end
113 %% Critical flow conditions
114 if omega>1
115 nc=0.55+0.217*log(omega)-0.046*((log(omega))^2)
116 +0.004*((log(omega))^3);
117 elseif omega ≤ 1
118 nc=(1+(1.0446-0.0093431*omega^0.5)*omega
119 ^(-0.56261))^(-0.70356+0.014685*log(omega));
120 end
272
L.2 SRV 273
121 if P2<(nc*P1)
122 Gomega=N3*nc*sqrt((P1*100000)/(v_1*omega));
123 %%% Note critical condition
124 crit_cond=1;
125 elseif P2 ≥ (nc*P1)
126 Gomega=N3*(sqrt(-2*(omega*log(n2)+(omega-1)
127 *(1-n2)))/(omega*((1/n2)-1)+1))*sqrt((P1*100000)/v_1);
128 %%% Note subcritical condition
129 crit_cond=0;
130 end
131 %% %%%%%%%%%%%% HNE-DS %%%%%%%%%%%%%%
132 if cond==1
133 N=0;
134 elseif cond==0
135 %%% for safety valves, control valve (full lift) --> a=2/5
136 % N=(x+Cp*T*(P1*100000)*((v_gl1/(hgl1^2))*log(1/nc)))^(2/5);
137 %%% for orifices, control valves, short nozzles --> a=3/5
138 % N=(x+Cp*T*(P1*100000)*((v_gl1/(hgl1^2))*log(1/nc)))^(3/5);
139 if cond_transp==3
140 N=(x+Cp*T*(P1*100000)*((v_gl1/(hgl1^2))*log(1/nc)))^(3/5);
141 % N=(x+Cp*T*(P1*100000)*((v_gl1/(hgl1^2))*log(1/nc)))
142 % ^(-0.0003*(P1-0.9)^2 - 0.0038*(P1-0.9) + 0.6625);
143 % N=(x+Cp*T*(P1*100000)*((v_gl1/(hgl1^2))*log(1/nc)))^(0.5);
144 elseif cond_transp==2
145 N=(x+Cp*T*(P1*100000)*((v_gl1/(hgl1^2))*log(1/nc)))^(2/5);
146 % N=(x+Cp*T*(P1*100000)*((v_gl1/(hgl1^2))*log(1/nc)))
147 % ^(0.0273*(P1-P2)+0.3527);
148 % N=(x+Cp*T*(P1*100000)*((v_gl1/(hgl1^2))*log(1/nc)))^(0.3);
149 end
150 end
151 omegaDS=((x*v_g1)/(v_1*k))+(((N2*Cp*T*(P1*100000))/v_1)
152 *((v_gl1/hgl1)^2)*N);
153 %% %%%%%%%%%%% API520 %%%%%%%%%%%%%%%
154 %%% Critical
155 if P2<(nc*P1)
156 Kdg=0.91398;
157 %%% Subcritical
159 Kdg=0.96028;
160 end
161 Kdl=0.734;
162 Kb=1; Kw=1; Kv=1; M=0.032; R=287.0028305; Z=1;
163 G_g=Kdg*Kb*(P1*100000)*sqrt(M/(R*T*Z))*sqrt(k*(2/(k+1))
164 ^((k+1)/(k-1)));
165 G_l=Kdl*Kw*Kv*sqrt(2*rho_l*((P1*100000)-(P2*100000)));
166 GAPI=((x/G_g)+((1-x)/G_l))^(-1);
167 %% Critical flow conditions
168 if omegaDS>1
169 nc=0.55+0.217*log(omegaDS)-0.046*((log(omegaDS))^2)
170 +0.004*((log(omegaDS))^3);
171 elseif omegaDS ≤ 1
172 nc=(1+(1.0446-0.0093431*omegaDS^0.5)*omegaDS^(-0.56261))
173 ^(-0.70356+0.014685*log(omegaDS));
174 end
175 if P2<(nc*P1)
273
274 L Matlab codes
176 G_HNEDS=N3*nc*sqrt((P1*100000)/(v_1*omegaDS));
177 crit_cond=1;
179 G_HNEDS=N3*(sqrt(-2*(omegaDS*log(n2)+(omegaDS-1)*(1-n2)))
180 /(omegaDS*((1/n2)-1)+1))*sqrt((P1*100000)/v_1);
181 crit_cond=0;
182 end
183 %% Results and plots
184 %%%Initinalize with zeros
185 Results(i,:)=[P1_rel,x,F,L,Gt1,Gomega,G_HNEDS,GAPI,
186 cond_transp,rho_cor,crit_cond];
187 % Results(i,:)=[P1_rel,x,F,L,Gt1,Gomega,G_HNEDS,
188 % GAPI,cond_transp,rho_cor,crit_cond,omega,n2,nc];
189 end
190 close all;
191 hdr={'Pset[bar]','x [-]', 'F [N]', 'L [mm]',
192 'Gt1 [kg/m^2s]', 'Gomega [kg/m^2s]', 'G_HNEDS [kg/m^2s]',
193 'G_API [kg/m^2s]', 'Transparent-metallic [-]',
194 'Corrected density [kg/m^3]','Critical condition [-]'};
195 % hdr={'Pset[bar]','x [-]', 'F [N]', 'L [mm]', 'Gt1 [kg/m^2s]',
196 % 'Gomega [kg/m^2s]', 'G_HNEDS [kg/m^2s]', 'G_API [kg/m^2s]',
197 % 'Transparent-metallic [-]''Corrected density [kg/m^3]',
198 % 'Critical condition [-]','omega','n2','nc'};
199 txt=sprintf('%s\t',hdr{:});
200 txt(end)='';
201 dlmwrite('Results.dat',txt,'');
202 dlmwrite('Results.dat', Results,'-append','delimiter',
203 '\t','precision',3);
204 %% Load and sort data for plotting %%%%
205 root='C:\Data\PhD\Safety_valve\All_Processing\Results';
206 ext='.dat';
209 % file=dlmread(filename,'',[1 0 170 9]);
211 fclose(fid);
212 %%% Keep only full lift results L=7.2 mm
213 file2=file(find(file(:,4)>7),:);
214 %%% Various openings results
215 % file2=file(find(file(:,4)==1),:);
216 Sorted_data=sortrows(file2,1);
217 %% All data plot
218 Res_P=Sorted_data(:,1);
219 Res_x=Sorted_data(:,2);
220 Res_F=Sorted_data(:,3);
221 Res_L=Sorted_data(:,4);
222 Res_Gt1=Sorted_data(:,5);
223 Res_Gomega=Sorted_data(:,6);
224 Res_GHNE_DS=Sorted_data(:,7);
225 Res_GAPI=Sorted_data(:,8);
226 Res_cond_transp=Sorted_data(:,9);
227 figure
228 semilogx(Res_x,Res_Gt1,'marker','o','markerfacecolor', 'r',
229 'markeredgecolor','k','linestyle','none','MarkerSize',11)
230 hold on
274
L.2 SRV 275
231 semilogx(Res_x,Res_Gomega,'marker','p','markerfacecolor',
232 'b','markeredgecolor','k','linestyle','none','MarkerSize',11)
233 semilogx(Res_x,Res_GHNE_DS,'marker','d','markerfacecolor',
234 'm','markeredgecolor','k','linestyle','none','MarkerSize',11)
235 semilogx(Res_x,Res_GAPI,'marker','s','markerfacecolor', 'k',
236 'markeredgecolor','r','linestyle','none','MarkerSize',11)
237 xlabel('\bfMass quality [-]','FontSize',16)
238 ylabel('\bfMass flux [kg/m^{2}s]','FontSize',16)
239 title('\bfMass flux for different pressures-experimental-models',
240 'FontSize',16)
241 grid
242 h_legend = legend('Experimental','Omega','HNE-DS','API 520',...
243 'Location','NorthEastOutside','FontSize',16);
244 set(h_legend,'FontSize',20)
245 %% Plot by Pressure
246 %%%%% Sort results by pressure
247 % %%%Transparent-metallic
248 % for j=1:170
249 % for cond_transp=2
250 % file(j,10)=4;
251 % end
252 % end
253 % file(j,10)=('Metallic');
254 % file(j,10)=('Transparent');
255 %%% Transparent
256 P_1_5_T=Sorted_data(find(Sorted_data(:,1)
257 <1.64&Sorted_data(:,1)>0.8),:);
258 P_3_T=Sorted_data(find(Sorted_data(:,1)
259 <2.85&Sorted_data(:,1)>2),:);
260 %%% Metallic
261 P_1_5_M=Sorted_data(find(Sorted_data(:,1)
262 <1.9&Sorted_data(:,1)>1.6),:);
263 P_3_M=Sorted_data(find(Sorted_data(:,1)
264 <3.2&Sorted_data(:,1)>2.9),:);
266 <5.1&Sorted_data(:,1)>4.9),:);
268 <6.1&Sorted_data(:,1)>5.9),:);
270 <7.1&Sorted_data(:,1)>6.9),:);
272 <9.1&Sorted_data(:,1)>8.9),:);
273 %% P=1.5 bar-transparent
274 Res_P_1_5_T=P_1_5_T(:,1);
275 Res_x_1_5_T=P_1_5_T(:,2);
276 Res_F_1_5_T=P_1_5_T(:,3);
277 Res_L_1_5_T=P_1_5_T(:,4);
278 Res_Gt1_1_5_T=P_1_5_T(:,5);
279 Res_Gomega_1_5_T=P_1_5_T(:,6);
280 Res_GHNE_DS_1_5_T=P_1_5_T(:,7);
281 Res_GAPI_1_5_T=P_1_5_T(:,8);
282 Res_Gcond_transp_1_5_T=P_1_5_T(:,9);
283 figure
284 semilogx(Res_x_1_5_T,Res_Gt1_1_5_T,'marker','o',
285 'markerfacecolor', 'r','markeredgecolor','k',
275
276 L Matlab codes
286 'linestyle','none','MarkerSize',11)
287 hold on
288 semilogx(Res_x_1_5_T,Res_Gomega_1_5_T,'marker','p',
289 'markerfacecolor', 'b','markeredgecolor','k',
290 'linestyle','-','MarkerSize',11)
291 semilogx(Res_x_1_5_T,Res_GHNE_DS_1_5_T,'marker','d',
292 'markerfacecolor', 'm','markeredgecolor','k',
294 semilogx(Res_x_1_5_T,Res_GAPI_1_5_T,'marker','s',
295 'markerfacecolor', 'k','markeredgecolor','r',
299 title('\bfP=1.5 bar-transparent valve','FontSize',16)
300 grid
301 h_legend = legend('Experimental','Omega','HNE-DS',
302 'API 520',...
305 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
306 %%% P=3 bar-transparent
307 Res_P_3_T=P_3_T(:,1);
308 Res_x_3_T=P_3_T(:,2);
309 Res_F_3_T=P_3_T(:,3);
310 Res_L_3_T=P_3_T(:,4);
311 Res_Gt1_3_T=P_3_T(:,5);
312 Res_Gomega_3_T=P_3_T(:,6);
313 Res_GHNE_DS_3_T=P_3_T(:,7);
314 Res_GAPI_3_T=P_3_T(:,8);
315 Res_Gcond_transp_3_T=P_3_T(:,9);
316 Res_rho_cor_cond_transp_3_T=P_3_T(:,10);
317 figure
318 semilogx(Res_x_3_T,Res_Gt1_3_T,'marker','o','markerfacecolor',
319 'r','markeredgecolor','k','linestyle','none','MarkerSize',11)
320 hold on
321 semilogx(Res_x_3_T,Res_Gomega_3_T,'marker','p','markerfacecolor',
322 'b','markeredgecolor','k','linestyle','-','MarkerSize',11)
323 semilogx(Res_x_3_T,Res_GHNE_DS_3_T,'marker','d','markerfacecolor',
324 'm','markeredgecolor','k','linestyle','-','MarkerSize',11)
325 semilogx(Res_x_3_T,Res_GAPI_3_T,'marker','s','markerfacecolor',
326 'k','markeredgecolor','r','linestyle','-','MarkerSize',11)
329 title('\bfP=3 bar-transparent valve','FontSize',16)
330 grid
334 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
335 %% P=1.5 bar-metallic
336 %%% Remark --> points taken from averages
337 %% --> abberating points
338 Res_P_1_5_M=P_1_5_M(:,1);
339 Res_x_1_5_M=P_1_5_M(:,2);
340 Res_F_1_5_M=P_1_5_M(:,3);
276
L.2 SRV 277
341 Res_L_1_5_M=P_1_5_M(:,4);
342 Res_Gt1_1_5_M=P_1_5_M(:,5);
343 Res_Gomega_1_5_M=P_1_5_M(:,6);
344 Res_GHNE_DS_1_5_M=P_1_5_M(:,7);
345 Res_GAPI_1_5_M=P_1_5_M(:,8);
346 Res_Gcond_transp_1_5_M=P_1_5_M(:,9);
347 %%%% Sort data not to have abberating points-continuous line
348 Res_x_1_5_M_sort=sort(Res_x_1_5_M);
349 Res_Gt1_1_5_M_sort=sort(Res_Gt1_1_5_M,'descend');
350 Res_Gomega_1_5_M_sort=sort(Res_Gomega_1_5_M,'descend');
351 Res_GHNE_DS_1_5_M_sort=sort(Res_GHNE_DS_1_5_M,'descend');
352 Res_GAPI_1_5_M_sort=sort(Res_GAPI_1_5_M,'descend');
353 %%%% plot correct
354 figure
355 semilogx(Res_x_1_5_M_sort,Res_Gt1_1_5_M_sort,'marker','o',
356 'markerfacecolor', 'r','markeredgecolor','k','linestyle','none',
357 'MarkerSize',11)
358 hold on
359 semilogx(Res_x_1_5_M_sort,Res_Gomega_1_5_M_sort,'marker','p',
360 'markerfacecolor', 'b','markeredgecolor','k','linestyle','-',
362 semilogx(Res_x_1_5_M_sort,Res_GHNE_DS_1_5_M_sort,'marker','d',
363 'markerfacecolor', 'm','markeredgecolor','k','linestyle','-',
365 semilogx(Res_x_1_5_M_sort,Res_GAPI_1_5_M_sort,'marker','s',
366 'markerfacecolor', 'k','markeredgecolor','r','linestyle','-',
370 title('\bfP=1.5 bar-metallic valve','FontSize',16)
371 grid
373 'API 520',...
376 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
377 %%% P=3 bar-metallic
378 Res_P_3_M=P_3_M(:,1);
379 Res_x_3_M=P_3_M(:,2);
380 Res_F_3_M=P_3_M(:,3);
381 Res_L_3_M=P_3_M(:,4);
382 Res_Gt1_3_M=P_3_M(:,5);
383 Res_Gomega_3_M=P_3_M(:,6);
384 Res_GHNE_DS_3_M=P_3_M(:,7);
385 Res_GAPI_3_M=P_3_M(:,8);
386 Res_Gcond_transp_3_M=P_3_M(:,9);
387 Res_rho_cor_cond_transp_3_M=P_3_M(:,10);
389 Res_x_3_M_sort=sort(Res_x_3_M);
390 Res_Gt1_3_M_sort=sort(Res_Gt1_3_M,'descend');
391 Res_Gomega_3_M_sort=sort(Res_Gomega_3_M,'descend');
392 Res_GHNE_DS_3_M_sort=sort(Res_GHNE_DS_3_M,'descend');
393 Res_GAPI_3_M_sort=sort(Res_GAPI_3_M,'descend');
395 figure
277
278 L Matlab codes
396 semilogx(Res_x_3_M_sort,Res_Gt1_3_M_sort,'marker','o',
397 'markerfacecolor', 'r','markeredgecolor','k','linestyle',
398 'none','MarkerSize',11)
399 hold on
400 semilogx(Res_x_3_M_sort,Res_Gomega_3_M_sort,'marker','p',
401 'markerfacecolor', 'b','markeredgecolor','k','linestyle',
402 '-','MarkerSize',11)
403 semilogx(Res_x_3_M_sort,Res_GHNE_DS_3_M_sort,'marker','d',
404 'markerfacecolor', 'm','markeredgecolor','k','linestyle',
406 semilogx(Res_x_3_M_sort,Res_GAPI_3_M_sort,'marker','s',
407 'markerfacecolor', 'k','markeredgecolor','r','linestyle',
411 title('\bfP=3 bar-metallic valve','FontSize',16)
412 grid
414 'API 520',...
417 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
419 Res_P_5_M=P_5_M(:,1);
420 Res_x_5_M=P_5_M(:,2);
421 Res_F_5_M=P_5_M(:,3);
422 Res_L_5_M=P_5_M(:,4);
423 Res_Gt1_5_M=P_5_M(:,5);
425 Res_GHNE_DS_5_M=P_5_M(:,7);
426 Res_GAPI_5_M=P_5_M(:,8);
435 figure
439 hold on
278
L.2 SRV 279

452 grid
454 'API 520',...
457 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
459 Res_P_6_M=P_6_M(:,1);
460 Res_x_6_M=P_6_M(:,2);
461 Res_F_6_M=P_6_M(:,3);
462 Res_L_6_M=P_6_M(:,4);
463 Res_Gt1_6_M=P_6_M(:,5);
465 Res_GHNE_DS_6_M=P_6_M(:,7);
466 Res_GAPI_6_M=P_6_M(:,8);
475 figure
479 hold on
492 grid
494 'API 520',...
497 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
499 Res_P_7_M=P_7_M(:,1);
500 Res_x_7_M=P_7_M(:,2);
501 Res_F_7_M=P_7_M(:,3);
502 Res_L_7_M=P_7_M(:,4);
503 Res_Gt1_7_M=P_7_M(:,5);
505 Res_GHNE_DS_7_M=P_7_M(:,7);
279
280 L Matlab codes
506 Res_GAPI_7_M=P_7_M(:,8);
515 figure
519 hold on
532 grid
534 'API 520',...
537 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
539 Res_P_9_M=P_9_M(:,1);
540 Res_x_9_M=P_9_M(:,2);
541 Res_F_9_M=P_9_M(:,3);
542 Res_L_9_M=P_9_M(:,4);
543 Res_Gt1_9_M=P_9_M(:,5);
545 Res_GHNE_DS_9_M=P_9_M(:,7);
546 Res_GAPI_9_M=P_9_M(:,8);
555 figure
559 hold on
280
L.2 SRV 281

572 grid
574 'API 520',...
577 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
578 %% Subplots
579 figure
580 subplot(2,2,1);
584 hold on
596 title('\bfP=0.3 MPa-metallic valve','FontSize',16)
597 grid
599 'API 520',...
602 subplot(2,2,2);
606 hold on
281
282 L Matlab codes

619 grid
621 'API 520',...
624 subplot(2,2,3);
628 hold on
641 grid
643 'API 520',...
646 subplot(2,2,4);
650 hold on
651 semilogx(Res_x_9_M_sort,Res_Gomega_9_M_sort,'-bp','marker',
652 'p','markerfacecolor', 'b','markeredgecolor','k','linestyle',
654 semilogx(Res_x_9_M_sort,Res_GHNE_DS_9_M_sort,'-md','marker',
655 'd','markerfacecolor', 'm','markeredgecolor','k','linestyle',
657 semilogx(Res_x_9_M_sort,Res_GAPI_9_M_sort,'-ks','marker','s',
663 grid
667 %%%%%%%%% TEST comparison transaprent-metallic 3 bar %%%%%%%%%%
668 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
669 % DP_M=3.1*100000;
670 DP_M=4*100000;
282
L.2 SRV 283
671 Norm_factor_M=sqrt(1000*DP_M);
672 % Norm_factor_M=sqrt(Res_rho_cor_cond_transp_3_M*DP_M);
673 Res_Gt1_3_M_norm_test=Res_Gt1_3_M/Norm_factor_M;
674 Res_Gt1_3_M_norm=Res_Gt1_3_M_norm_test(:,1);
675 % DP_T=2.1*100000;
676 DP_T=4*100000;
677 Norm_factor_T=sqrt(1000*DP_T);
678 % Norm_factor_T=sqrt(Res_rho_cor_cond_transp_3_T*DP_T);
679 Res_Gt1_3_T_norm_test=Res_Gt1_3_T/Norm_factor_T;
680 % Res_Gt1_3_T_norm=Res_Gt1_3_T_norm_test(:,8);
681 Res_Gt1_3_T_norm=Res_Gt1_3_T_norm_test(:,1);
682 figure
683 semilogx(Res_x_3_M,Res_Gt1_3_M_norm,'marker','o','markerfacecolor',
684 'r','markeredgecolor','k','linestyle','none','MarkerSize',11)
685 hold on
686 semilogx(Res_x_3_T,Res_Gt1_3_T_norm,'marker','s','markerfacecolor',
687 'k','markeredgecolor','r','linestyle','none','MarkerSize',11)
689 ylabel('\bfG* [-]','FontSize',16)
690 title('\bfDimensionless mass flux (G^{*}=G/(\DeltaP*\rho^{*})^{0.5})
691 -metallic-transparent valve-experimental comparison','FontSize',16)
692 grid
693 h_legend = legend('Experimental-metallic',
694 'Experimental-transparent',...
283

Experimental Study and Modeling of Single - and Two-Phase Flow in Singular Geometries and Safety Relief Valves

Încărcat de

Informații document

Titlu original

Drepturi de autor

Formate disponibile

Partajați acest document

Partajați sau inserați document

Opțiuni de partajare

Vi se pare util acest document?

Este necorespunzător acest conținut?

Drepturi de autor:

Formate disponibile

Experimental Study and Modeling of Single - and Two-Phase Flow in Singular Geometries and Safety Relief Valves

Încărcat de

Drepturi de autor:

Formate disponibile

Université Libre de Bruxelles

Faculté des sciences appliquées

Doctorate of Science in Engineering

Experimental study and modeling of

prepared at Bruxelles, Belgium

Jury: Jean-Marie Buchlin, Patrick Rambaud, Saïd Chabane, Jürgen

"A man goes to knowledge as he goes to war:

θα ήθελα, τέλος, να ευχαριστήσω την οικογένειά μου που με έχει στηρίξει

List of figures vii

List of symbols xvii

2 Theoretical background-Literature review 15

2.3.3.1 Expansion geometries . . . . . . . . . . . . . . . . . . 31

3.3.1 Flow meters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85

III Geometrical singularities 93

5 Contraction singularities 127

IV Safety Relief Valve 141

6.4.3.1 CFD results . . . . . . . . . . . . . . . . . . . . . . . 161

V Conclusions and future recommendations 181

Appendix A Uncertainty analysis 199

Appendix B Design of experimental facilities 203

Appendix C Disasters caused by PRV failures 209

Appendix D PED Annex II diagrams 213

Appendix E PRV orifice sizes 217

Appendix F Vapor-liquid flow 219

Appendix G Additional flow regime charts 221

Appendix H Pressure drop in straight pipe SP flow 229

Appendix I Straight pipe two-phase flow models 233

Appendix J Single-phase pressure change in singularities 243

Appendix K Pressure drop database 249

Appendix L Matlab codes 265

1.1 Different types of pipe fittings-geometrical singularities (SS Engineers &

2.1 Flow patterns in horizontal flow in a pipe. Adapted by Weisman (1983). . 21

3.1 A) Progressive expansion of different opening angles-reattachment length

4.4 Dimensionless singular pressure change ΦL versus volumetric quality.

B.1 Detailed optical probe test section. . . . . . . . . . . . . . . . . . . . . . 204

C.1 Three Mile Island nuclear power plant. . . . . . . . . . . . . . . . . . . 210

D.1 Diagrams 1-4 of Annex II in PED . . . . . . . . . . . . . . . . . . . . . 214

F.1 Vertical and horizontal evaporation-condensation in a tube (from Collier

H.1 Moody (1944) diagram. . . . . . . . . . . . . . . . . . . . . . . . . . . 231

I.3 Acceleration multiplier r2 (left), gravitational multiplier r4 (right up) and

2.1 Classification of relief devices Tyco (2008). . . . . . . . . . . . . . . . . 37

3.1 Presentation of the four main experimental facilities. . . . . . . . . . . . 65

5.1 Proposed correlation for different contraction singularities-regression pa-

6.1 Flow conditions for SRV tests. . . . . . . . . . . . . . . . . . . . . . . . 161

E.1 Relief Valve Orifice Sizes . . . . . . . . . . . . . . . . . . . . . . . . . . 217

J.2 Pressure changes in abrupt contraction and expansion geometries in in-

K.1 Results in sudden expansion σ=0.43. . . . . . . . . . . . . . . . . . . . 249

Q Volumetric flow rate [l/s] . . . . . . . . . . . . . . . . . . . . . . 15

CP Specific heat capacity of a gas for constant pressure [-] . . . . . . 40

T f light Transient time between primary and secondary sensor [ms] . . . . 89

D Large pipe diameter [m] . . . . . . . . . . . . . . . . . . . . . . 229

LM Lockhart and Martinelli (1949) parameter

ECT Electrical Capacitance Tomography

in the case of their homogeneous mixing this ratio is equal to unity.

1.1.2 Two-phase flow occurring in industrial devices and safety valves

Figure 1.2: Detailed sketch of nuclear reactor core.

1.2 Objectives-Achievements / Motivation

Therefore, the main objective of this study is to determine whether a two-

I Measurements of flowforce in SP and TP flow revealing the compressible

Safety valve Qm,out

Figure 1.5: Cutaway of PRV. Figure 1.6: Schematic of the problem.

Figure 1.7: Deepwater Horizon oil spill accident (2010).

The Pressure Equipment Directive (PED)