33927175

19 1727310 5
R E F E R E N C E O N L>
UNIVERSITY OF LONDON THESIS
Degree P ^ D > Year T }jD O 6 Name of Author }

U ^ u ry
C O P Y R IG H T
This is a thesis accepted for a Higher Degree of the University of London. It is an
unpublished typescript and the copyright is held by the author. All persons consulting
the thesis must read and abide by the Copyright Declaration below.
C O P Y R IG H T D E C L A R A T IO N
I recognise that the copyright of the above-described thesis rests with the author and
that no quotation from it or information derived from it may be published without the
prior written consent of the author.
LO ANS
Theses may not be lent to individuals, but the Senate House Library may lend a copy
to approved libraries within the United Kingdom, for consultation solely on the
premises of those libraries. Application should be made to: Inter-Library Loans,
Senate House Library, Senate House, Malet Street, London WC1E 7HU.
R E P R O D U C T IO N
University of London theses may not be reproduced without explicit written
permission from the Senate House Library. Enquiries should be addressed to the
Theses Section of the Library. Regulations concerning reproduction vary according
to the date of acceptance of the thesis and are listed below as guidelines.
A. Before 1962. Permission granted only upon the prior written consent of the
author. (The Senate House Library will provide addresses where possible).
B. 1962- 1974. In many cases the author has agreed to permit copying upon
completion of a Copyright Declaration.
C. 1975 - 1988. Most theses may be copied upon completion of a Copyright

Declaration.
D. 1989 onwards. Most theses may be copied.
This thesis comes within category D.
□ This copy has been deposited in the Library of
| I This copy has been deposited in the Senate House Library, Senate House,
LJ/ Malet Street, London WC1E 7HU.
C :\D ocum ents and Settings\lproctor\Local SettingsYTemporary internet Files\OLK8\Copyright - th esis (2).doc
METHODS FOR ASSESSING ENVIRONMENTAL, SAFETY AND PERFORMANCE
OF ALTERNATIVE REFRIGERANTS
Daniel Henry Colbourne
A thesis submitted for the degree of Doctor of Philosophy in the Department of Mechanical
Engineering, University College London
June 2006
UMI Number: U592706
All rights reserved
INFO RM ATIO N TO ALL USERS

The quality of this reproduction is dependent upon the quality of the copy submitted.
In the unlikely event that the author did not send a complete manuscript
and there are missing pages, these will be noted. Also, if material had to be removed,
a note will indicate the deletion.
Dissertation Publishing
UMI U592706
Published by ProQuest LLC 2013. Copyright in the Dissertation held by the Author.
Microform Edition © ProQuest LLC.
All rights reserved. This work is protected against
unauthorized copying under Title 17, United States Code.
ProQuest LLC
789 East Eisenhower Parkway
P.O. Box 1346
Ann Arbor, Ml 48106-1346
DECLARATION
This declares that the content presented in this thesis is the work o f the author.
Daniel Henry Colbourne.
Date: 20thJune 2006
2
A B STR A CT
This thesis addresses the general subject o f implementation o f refrigerants that are alternatives for
ozone depleting substances. In particular it covers three topics, which are environmental
assessment, system performance and safety aspects related to flammability. The first Part
examines the current means o f assessing the global warming impact o f systems, and proposes an
improved approach. The second Part presents a new balancing technique for designing systems
that use refrigerants with a temperature glide, as an advancement over the conventional technique
for pure refrigerants. The third Part derives a quantitative risk assessment model to be employed
for evaluating the safety o f hazard o f flammable refrigerants.
The concept o f ‘T otal Equivalent Warming Impact” is introduced in Part 1, which is currently
used by manufacturers, installers and end-users o f refrigerating systems to evaluate the
contribution to climate change resulting from energy consumption and refrigerant leakage. It was
found that existing methods rely on many assumptions, so a thorough approach was proposed to
assess the relevance o f the assumptions and providing ways o f avoiding them. An alternative
measure for interpretation o f consequences o f greenhouse gas emissions is also included. Part 2
investigated the existing approach for rating and balancing system components, which w ere found
to be incompatible when applied to zeotropic refrigerants. Differences in the performance of
components using pure and zeotropic refrigerants were identified, and so a new method for
component rating and an advanced system balancing technique were developed based on the
properties specific to zeotropes. Within Part 3, a framework for calculating ignition frequencies
and consequences o f a release of hydrocarbon refrigerant is presented. A major element o f these
calculations is the evaluation o f certain quantities o f the flammable mixture, and the model
developed for this purpose w as based on the results from a series o f gas dispersion experiments.
3
ACKNOWLEDGEMENTS
The Author would like to express appreciation to many people. First, my tutor, Dr. K. 0 . Suen,
for his incredible tolerance and stamina, his endless enthusiasm and encouragement, and for
making available his inspirational way o f thinking, understanding and addressing all matters
technical. Also, my parents, Jill and Alan, for their encouragement, support, motivation, financial
help, and o f course my creation and subsequent existence. Notwithstanding, my brother Rupert,
particularly for the invaluable help with coding and moreover the highly amusing observations
about things irrelevant to this work. Many colleagues from my previous employer Calor Gas
Ltd., including Paul Blacklock and Loretta Powell for their support and encouragement, but in
particular Dr. T ern ’ Ritter for his keen interest, enthusiastic discussions, and moral support.
Lastly, thanks to a formidable set o f friends whom for years have endured endless lectures and
discussions whilst having a profound ability to convincingly express fascination in the subjects
detailed herein.
4
TABLE OF CONTENTS
LIST OF TABLES 11
LIST OF FIGURES 13
GENERAL INTRODUCTION 20
PART 1: METHODS FOR EVALUATING THE GLOBAL WARMING IMPACT OF

REFRIGERATING SYSTEMS 23
NOMENCLATURE 24
SECTION 1: INTRODUCTION 29
1.1 Environmental burden o f refrigerants 29
1.2 Literature review 30
1.2 .1 Existing methodologies for evaluating environmental impacts 30
1.2 .2 Survey o f comparative studies 32
1.2.3 Measures o f climatic impacts 34
1.3 Formulation o f objectives 37
1.3.1 Summary o f findings 38
1.3.2 Aims and objectives 39
SECTION 2: DEVELOPMENT OF EMISSIONS MODEL 41
2.1 Modelling approach 41
2.2 Start o f life stage 43
2.2.1 Refrigerant and component manufacture 43
2.2.2 Construction, assembly and installation 49
2.3 In-use stage 50
2.3.1 System refrigerant emissions 51
2.3 .2 System energy use emissions 54
2.3.3 Emissions from servicing and maintenance activities 70
2.4 End of life stage 73
2.4.1 Dismantling 74
2 .4 .2 Disposal of refrigerant and equipment 75
SECTION 3: CLIMATE IMPACTS 76
3 .1 Options for evaluating climate change impact 76
3.2 Global warming potential 77
3.3 Global temperature change potential 78
3.4 Total lifetime temperature impact (TLTI) 79
5
3 .5 Global damage index (GDI) 80
SECTION 4: RESULTS AND DISCUSSION 82
4.1 Implementation of model 82
4.2 Breakdown and uncertainty o f emissions 82
4 .2 .1 Breakdown o f emission sources 82
4.2.2 Uncertainties of emissions 85
4.3 Sensitivity to transient influences 86
4.3.1 External conditions 87
4.3.2 Ageing 87
4.3.3 Power station emissions 88
4.4 Influences of equipment design 89
4.4.1 Pipework and connections 89
4.4.2 Heat exchanger size 90
4.4.3 Compressor size 93
4.4.4 Refrigerant selection 93
4.5 Effects o f material/refrigerant handling and operations 98
4.5.1 Refrigerant handling practices and recycling 99
4.5.2 Condenser and system cleaning 101
4 .5 .3 Leak checks and repair 102
4.6 Comparison with BRA method 103
4.7 Interpretation o f global warming 105
4 .8 Use and advantages o f methodology' 109
SECTION 5: CONCLUSIONS 111
5.1 Summary of w ork 111
5.2 Key findings 111
5.3 Limitations and further work 115
REFERENCES 117
APPENDICES 124
Appendix 1: Compilation of sources found to affect emissions 124
PART 2: DEVELOPING A NEW SYSTEM BALANCING TECHNIQUE FOR MIXTURE

REFRIGERANTS 125
NOMENCLATURE 126
1.1 New7refrigerants 133
6
1.2 Overview o f refrigeration component-system chain 133
1.3.1 Performance testing, rating, selection and balancing 136
1.3.2 Implications o f zeotropic refrigerants 140
1.4 Formulation of objectives 141
1.4 .1 Summary o f findings 142
SECTION 2: IMPLICATIONS ON COMPONENT PERFORMANCE 145
2.1 Initial consideration of impact o f glide on component performance 145
2.2 Compressor 146
2.3 Expansion device 147
2.4 Condenser and evaporator 148
2.4.1 Analysis o f heat exchanger rating with pure refrigerants 149
2.4.2 Analysis o f heat exchanger for refrigerants with temperature glide 151
2.4.3 Implications o f temperature glide behaviour 157
2.5 Findings relevant to component-system chain and additional considerations 166
SECTION 3: CHARACTERISATION OF COMPONENTS WITH ZEOTROPES 168
3.1 Approach for improved characterisation 168
3.2 Simulation o f compressor performance 168
3.2.1 General modelling approach 168
3.2.2 Semi-empirical model 169
3 .2 .3 Determination o f constants 170
3.3 Simulation o f condenser and evaporator performance 171
3.3.1 General modelling approach 171
3.3.2 Condenser heat transfer and pressure loss 176
3.3.3 Evaporator heat transfer and pressure loss 179
3.3.4 Single-phase heat transfer coefficient 185
3.4 Observations from component performance simulation 187
3.5 Proposed characterisation methods 196
3.5.1 Compressor 196
3.5.2 Expansion device 198
3.5.3 Heat exchangers 198
SECTION 4: SYSTEM BALANCING 208
4.1 Introduction 208
4.2 Balance technique theory 208
7
4.3 Balance technique for zeotropic refrigerants 210
4.4 New balance technique for zeotrope with desuperheat and pressure loss 213
4.5 Clarification o f new procedures 222
SECTION 5: ILLUSTRATIVE EXAMPLE OF NEW TECHNIQUES 223
5.1 Introduction 223
5.2 Generation o f component data 223
5.3 Illustration o f worked example 225
5.4 Comparison o f results 230
SECTION 6: FINAL SUMMARY AND CONCLUSION 233
6.1 Summary’ o f work 233
6.2 Key findings and originality 233
REFERENCES 237
APPENDICES 242
Appendix 1: Information on refrigerant mixtures 242
Appendix 2: Calculation o f HTF properties 245
Appendix 3: Overview o f new balancing technique 246
Appendix 4: Results o f worked examples for balancing 248
PART 3: APPRAISING THE FLAMMABILITY HAZARDS OF HYDROCARBON

REFRIGERANTS 249
NOMENCLATURE 250
1.1 Hydrocarbon refrigerants 260
1.1.1 Safety standards and codes 260
1.1.2 Risk assessment for refrigerating systems 261
1.2.1 Risk assessment o f flammable gas 262
1.2.2 Risk assessment o f flammable refrigerants 263
1.2.3 General gas dispersion in confined spaces 265
1.2.4 Releases into quiescent rooms 266
1.2.5 Releases into rooms with ventilation 269
1.3 Formulation o f objectives 272
1.3.1 Summary7o f findings 272
8
SECTION 2: GENERAL RISK ANALYSIS 276
2.1 Fundamental risk model 276
2.2 Context and risk acceptance criteria 280
2.3 Leakage frequency 283
2.4 Probability o f ignition 287
2.5 Time fractions 294
SECTION 3: DISPERSION OF REFRIGERANT LEAKS 299
3.1 Dispersion experiments 299
3.1.1 Experiment rationale 299
3.1.2 Apparatus set-up and experimental procedure 305
3.1.3 Interpretation o f Results 313
3.2 Development of correlation for floor level concentration 330
3.2.1 Maximum floor concentration from catastrophic release 331
3.2.2 Floor concentration from non-catastrophic releases 336
3.2.3 Verification o f correlations 339
3.3 Use o f correlations for design purposes 342
3.3.1 Maximum concentration from catastrophic leak 343
3.3.2 Maximum concentration from non-catastrophic leaks 343
3.3.3 Application to control of airflow 345
3.4 Development o f flammable volume/time models 347
3 .4.1 General modelling approach 347
3.4.2 Spreading velocity 352
3 .4.3 Descending plume model 355
3.4.4 Spreading plume model 358
3.4.5 Cloud decay model 362
3.4.6 Airflow models 363
3.4.7 Model verification 365
3 .4.8 Application 368
SECTION 4: CONSEQUENCE ANALYSIS 374
4 .1 Selection o f consequences 374
4.2 Thermal radiation model 375
4.3 Overpressure model 376
4.4 Application 382
SECTION 5: APPLICATION OF OVERALL MODEL 386
5.1 Uses for the QRA model 386
9
5.2 Evolution o f risk calculation 389
5.3 Parametric assessment o f input variables 395
5 .4 Example calculations for selected RAC equipment 401
5.5 General risk dependency on maximum concentration 405
SECTION 6: CONCLUSIONS 408
6.1 Summary o f work 408
6.2 Key findings and originality 408
REFERENCES 413
APPENDICES 421
Appendix 1: Data from published QRAs 421
Appendix 2: Data for dispersion experiments 424
Appendix 3: Validation data from other sources 427
Appendix 4. Properties o f flammable refrigerants 438
Appendix 5: Data for overpressure calculations 440
10
LIST OF TABLES
Table 1: GHG (refrigerant) data from IPCC/TEAP (2005) 36

Table 2: Manufacturing emissions o f refrigerants 45
Table 3: Component construction data 46
Table 4: Emissions resulting from raw material manufacture 48
Table 5: Emissions during construction, assembly and installation 49
Table 6: Refrigerant leak quantity and leak factor for specific components 52
Table 7: Comparison o f reported annual leak rates 53
Table 8: Constants for emissions factor projections 55
Table 9: Constants for equation (36) for two different compressors 63
Table 10: Constants for fouling equation (45) 67
Table 11: Studies on effects of refrigerant leakage 68
Table 12: Details o f servicing activities leading to refrigerant emissions 71
Table 13: Summary o f reported venting probabilities 72
Table 14: Reference input parameters and basic outputs 83
Table 15: Range o f emission contributions from different sources 85
Table 16: Effect of refrigerant leakage from different component assemblies 90
Table 17: Refrigerant comparison (a) with fixed system design 94
Table 18: Refrigerant comparison (b) with fixed system capacity at design conditions 96
Table 19: Refrigerant comparison (c) with fixed system COP at design conditions 97
Table 20: Input parameters for refrigerant handling practice 100
Table 21: Energy calculations according to BRA method 104
Table 22: Standard calculations according to BRA method 105
Table 23: Comparison o f climate impact for different systems 107
Table 24: Exhaustive list o f sources of emissions and influencing parameters 124
Table 25: Heat exchanger simulation input values 186
Table 26: Contribution o f TEV inlet temperature to factors affecting capacity 195
Table 27: Comparison o f capacity and temperature variables for balancing 213
Table 28: Comparison o f capacity and reference temperatures for balancing procedure 222
Table 29: Specific input data for calculation of fictitious rating data 224
Table 30: Summary o f usable data 235
Table 31: Commercial zeotropic mixtures and selected properties 242
Table 32: Constants for water property polynomial equation (265) 245
Table 33: Overview o f new balancing technique 246
11
Table 34: Comparison o f equilibrium-points for example balancing exercises 248
Table 35: Summary o f the ways that main parameters were determined in QRAs 274
Table 36: Maximum acceptable risk criteria 282
Table 37: Information on mechanical component failures 295
Table 38: Frequency o f mean monthly temperatures in West London 297
Table 39: Summary o f test parameters 300
Table 40: Information on RAC equipment 307
Table 41: Information on instrumentation used 308
Table 42: General test schedule 310
Table 43: Sample point location and representative volume 311
Table 44: Information on test parameters 329
Table 45: Summary o f influencing variables 332
Table 46: Constants/coefficients and indices for pi-groups and for the adjustment factor 335
Table 47: Range o f applicability for correlations 342
Table 48: Input details for dependence o f flammable volume-time on number of elements 365
Table 49: Reports o f incidents involving HC refrigerants 388
Table 50: Comparison with maximum acceptable risk criteria 395
Table 51: Reference values and variations 396
Table 52: Results o f parametric evaluation 397
Table 53: Reference values and variations 401
Table 54: QRA model results for servicing on four applications 404
Table 55: Summary o f QRA results for stationary' RAC equipment 421
Table 56: Leak frequencies used in previous studies 422
Table 57: Location of sensors and corresponding representative volume ratio 424
Table 58: Parameter values and principle results for CCL release tests 425
Table 59: Test variables and calculated results for Aloisi 427
Table 60: Test variables and calculated results for Clodic et al 428
Table 61: Test variables and calculated results for Jabbour et al, 2003 430
Table 62: Test variables and calculated results for Kataoka et al 434
Table 63: Flammable properties of hydrocarbons and selected refrigerants 438
Table 64: Data from overpressure measurements and model results 440
12
LIST OF FIGURES
Figure 1: Identification o f sources and types o f GHG emissions 42

Figure 2: Activities at start o f life stage causing emissions 43
Figure 3: Data for compressor and compressor oil mass 47
Figure 4: Data for condenser and evaporator mass 47
Figure 5: Data for pipe and valve mass 48
Figure 6: Overview o f in-use emission sources and influencing parameters 50
Figure 7: Proportion of refrigerant leaks by mass and number o f visits (Radford, 1998) 51
Figure 8: Historical and projected power station C 02 emission factor (DTI,1998, 2000) 54
Figure 9: Options for different types o f performance calculation 57
Figure 10: Changes in heat transfer of fouling over time 66
Figure 11: Comparison o f leak effects on COP 69
Figure 12: Activities producing emissions during servicing 70
Figure 13: Activities affecting emissions during end o f life process 74
Figure 14: Chain o f climate relevant measures 76
Figure 15: Illustration o f TLTI, GDI1 and GDI2 interpretation 80
Figure 16: Breakdown o f emissions from energy and refrigerant for different stages (based
on GWP( 100), In-use values for a single year only) 84
Figure 17: Variation o f contributions from different emission sources with sizeof
equipment 85
Figure 18: Variation in monthly emissions from energy consumption for different ambient
scenarios 86
Figure 19: Influence if various ageing effects on emissions from system energy’
consumption 87
Figure 20: Influence of change in power station emissions for electricity production 88
Figure 21: Effect of evaporator and condenser size on annual emissions 91
Figure 22: Influence of system sensitivity to charge size on annual emissions from energy’
consumption 91
Figure 23: Effect o f sensitivity to charge size on system performance 92
Figure 24: Effect o f compressor size on emissions from system operation 93
Figure 25: Comparison o f total emissions for different refrigerants 98
Figure 26: Comparison o f emissions for different service and handling practices 99
Figure 27: Influence o f condenser cleaning and system cleaning on emissions 100
Figure 28: Influence o f condenser and system cleaning on mean application temperature 101
13
Figure 29: Effect o f servicing repair and top-up frequency on emissions 102
Figure 30: Effect o f servicing repair and top-up frequency on mean application
temperature (design application: -20°C) 103
Figure 31: Comparison o f AGTP-time profile for high and low climate impact refrigerant 106
Figure 32: Climate impacts for different system characteristics relative to R22 reference
system 108
Figure 33: Climate impacts for different system characteristics relative to R22 and R290
TEWI 109
Figure 34: Stages in the component-system chain 134
Figure 35: Pressure enthalpy diagram for (a) pure refrigerant, and (b) zeotropic refrigerant 145
Figure 36: Graphical representation o f (a) condenser, and (b) evaporator characteristic for
pure refrigerant (where B is constant) 149
Figure 37: Temperature profiles o f counter-flow condenser for reducing condensing
temperature with pure refrigerant 151
Figure 38: Effect o f apparent capacity rate on effectiveness 153
Figure 39: Effect o f apparent capacity rate on basic rating 153
Figure 40: Effect o f temperature glide on Cmin. s and B in counter-flow7exchanger 154
Figure 41: Temperature profiles for counter-flow7condenser 155
Figure 42: Temperature profiles of parallel- and counter-flow’ condenser for reducing
condensing temperature with glide 15 5
Figure 43: Minimum approach temperature differences for condenser and evaporator 156
Figure 44: Graphical representation of (a) condenser and (b) evaporator characteristic with
zeotropic refrigerant 157
Figure 45: Variation o f temperature glide (a) with saturation temperature, and (b) with
vapour quality at atmospheric pressure for commercial refrigerant blends 158
Figure 46: Correlation for constants (a) Cgi and (b) Cg2 for equation (101) 159
Figure 47: Variation o f maximum and minimum glide index over (a) saturation
temperature, and (b) saturation pressure 160
Figure 48: Change in condenser basic rating with inlet temperature for fixed (hollow7data
point) and variable (filled data point) temperature glide 162
Figure 49: Effect o f different expansion device inlet temperature on (a) evaporator inlet
temperature, and (b) on evaporator characteristic 163
Figure 50: Variation in evaporator basic rating with glide size at different TxEvcin) 164
Figure 51: Variation o f evaporator capacity with glide size due to change in TEV inlet
temperature from 20°C to 60°C 164
14
Figure 52: Calculation sequence for condenser capacity 176
Figure 53: Calculation sequence for evaporator capacity 182
Figure 54: Counter-flow condenser, C 'r >C htf 187
Figure 55: Counter-flow condenser, C 'r - C htf 187
Figure 56: Counter-flow condenser with very small approach, C ’r « C htf 188
Figure 57: Counter-flow condenser with pure refrigerant and a small AT^n) 188
Figure 58: Characteristic line for condenser with pure and zeotropic refrigerant 189
Figure 59: Influence o f glide shape on temperature profiles for small ATa:,n) 189
Figure 60: Variation o f capacity7with ATc(m) for different no 190
Figure 61: Condenser temperature profile with single-phase heat transfer 191
Figure 62: Condenser characteristic affected by single-phase heat transfer 192
Figure 63: Effect o f pressure drop on refrigerant temperature in condenser 193
Figure 64: Effect of pressure loss on condenser characteristic (two-phase only) 194
Figure 65: Effect o f different inlet temperature on evaporator capacity 194
Figure 66: Effect of different inlet temperature on evaporator capacity7 195
Figure 67: Range of performance prediction methods for component rating 196
Figure 68: Graphical interpretation o f equation (239) 209
Figure 69: Graphical interpretation o f equation (247) 212
Figure 70: Refrigeration cycle with pressure loss and superheat 213
Figure 71: New condenser characteristic 215
Figure 72: Discharge superheat characteristic 215
Figure 73: Modified condenser characteristic 216
Figure 74: Compressor and condenser balancing 217
Figure 75: Evaporator characteristic 217
Figure 76: Condenser pressure loss characteristics 218
Figure 77: Converted condenser pressure loss characteristics 219
Figure 78: Modified evaporator characteristics 219
Figure 79: Balancing o f evaporator and compressor to form an evaporating unit 220
Figure 80: Evaporator balance point for condensing and evaporating unit 220
Figure 81: Condenser balance point for condensing and evaporating unit 221
Figure 82: Compressor capacity, against evaporating dew-point temperature for a series o f
condensing dew-point temperatures 225
Figure 83: Compressor discharge superheat, against condensing dew-point for a series of
evaporating dew-point temperatures 225
15
Figure 84: Condenser capacity, against condensing dew-point temperature for a series o f
discharge superheats 226
Figure 85: Condenser outlet dew-point temperature, against condenser inlet dew-point
temperature for a series of desuperheats 226
Figure 86: Condenser capacity against condensing dew-point temperature, for a series o f
evaporating temperatures 227
Figure 87: Condenser outlet dew-point temperature, against condensing inlet dew-point
temperature for a series of evaporating temperatures 227
Figure 88: Compressor characteristic superimposed on new condenser characteristic to
obtain condensing unit characteristic 227
Figure 89: Evaporator capacity against evaporating dew-point temperature for a series o f
condensing outlet dew-point temperatures 228
Figure 90: Evaporator characteristic corrected for condenser inlet 228
Figure 91: Compressor characteristic superimposed on new evaporator characteristic to
obtain evaporating unit characteristic 229
Figure 92: System balance condensing dew-point temperature 230
Figure 93: System balance evaporating dew-point temperature 230
Figure 94: Error in Qe and Qc at balance points with (a) small size compressor, and (b)
large size compressor for conventional and new techniques 231
Figure 95 (a) Layer formation in a quiescent room and, (b) concentration profiles 267
Figure 96: Positioning o f air inlet and outlet positions and release locations within a room 270
Figure 97: Sequence o f events leading to flammability hazard 276
Figure 98: Partial event tree illustrating variations o f conditions 278
Figure 99: Calculation sequence to determine adjustment for selected number of leak sizes 285
Figure 100: Venn diagram indicating occurrences necessary for ignition 289
Figure 101: Time fraction for different infiltration rate ranges 297
Figure 102: Schematic arrangement of the test chamber apparatus 306
Figure 103: Comparison o f maximum and mean maximum concentrations (r2 = 0.86) 312
Figure 104: Observations from smoke-only tests 313
Figure 105: Observations from smoke-C02 tests 314
Figure 106: Concentration-time profiles for reference test (no. 10) 315
Figure 107: Concentration-height profiles for reference test (no. 10) 315
Figure 108: Concentration-height profiles for small release mass 316
Figure 109: Concentration-height profiles for large release mass 316
Figure 110: Effect o f released mass on Qmax (r2 = 0.96) 316
16
Figure 111: Concentration-time profile for release with no airflow 317
Figure 112: Concentration profiles for low airflow rate (test no. 28) 318
Figure 113: Concentration profiles for high airflow’ rate (test no. 11) 318
Figure 114: Effect of AHU airflow rate on Cfmax (r2 = 0.85) 319
Figure 115: Concentration-height profiles for downward air discharge (test no 14.) 320
Figure 116: Concentration-height profiles for upward air discharge (test no 15.) 320
Figure 117: Effect of air discharge angle on Cfjnax (r2 = 0.82) 320
Figure 118: Concentration-height profiles resulting from thermal loads (test no. 43) 321
Figure 119: Effect o f post-ventilation period on Cfmax with 3.5 minutes release time. 322
Figure 120: Effect o f release time with 3.5 minute post-ventilation on Cfmax 322
Figure 121: Concentration-height profile for a 7 minute release (test no. 29) 323
Figure 122: Concentration-height profile for a 14 minute release (test no. 25) 323
Figure 123: Concentration-height profiles for low’ release height (test no. 20) 323
Figure 124: Concentration-height profiles for low release and discharge height (test no. 21) 323
Figure 125: Effect o f release height on Cfmax (r2 = 0.96) 324
Figure 126: Effect o f release height and unit height (Vhd) on Cfmax (r2 = 0.93) 324
Figure 127: Effect o f release location (horizontal distance from air disch.) on Cfmax (r2 =
0.93) 325
Figure 128: Configuration o f AHU airflow s showing discharge and return air paths 326
Figure 129: Configuration o f cabinet air flows 327
Figure 130: Concentration-height profiles for AHU #2 (test no. 32) 327
Figure 131: Concentration-height profiles for AHU #3 (test no. 36) 327
Figure 132: Concentration-height profiles for cabinet #1 (test no. 40) 328
Figure 133: Concentration-height profiles for cabinet #2 (test no. 41) 328
Figure 134.Comparison of measured room air speed (r2 = 0.92) 333
Figure 135: Correlation foffU (r2 = 0.88) 334
Figure 136: Correlation forn3 (r2 = 0.97) 334
Figure 137: Gradient o f floor concentration development with time 336
Figure 138: Correlation between n and mixing effectiveness (r2 = 0.74) 338
Figure 139: Comparison of calculated against measured concentrations 339
Figure 140: Comparison of correlation against measurements (overall r2 = 0.87) 340
Figure 141: Comparison of correlations against data from other studies (r2 = 0.78) 340
Figure 142: Change in concentration over time, with forced airflow ceased at td during
release 344
Figure 143: Effect of release time exceeding airflow time on final floor concentration 345
17
Figure 144: Change in concentration with initiation and termination o f airflow’ 346
Figure 145: Diagram showing basic concept of dispersion models 348
Figure 146: Flow chart for calculation sequence o f flammable quantities 350
Figure 147: Comparison o f plume spread correlations and measurements (r2 = 0.43) 353
Figure 148: Correlation o f friction factor and Reynolds no. for spread velocity (r2 = 0.67 353
Figure 149: Schematic diagram o f plume model 356
Figure 150: Illustration of room velocity profile 360
Figure 151: Dependence o f number o f radial elements on flammable volume-time 366
Figure 152: Comparison o f height-concentration profile at 120 s interval from model and
Test no. 9 ( no airflow) 367
Figure 153: Comparison o f height-concentration profile at 120 s interval from l model and
Test no. 28 (low airflow) 367
Figure 154: Comparison of concentrations for reference case (Test no. 10) 368
Figure 155: Comparison o f concentrations for low' release height (Test no. 20) 368
Figure 156: Comparison of concentrations for low release and air disch. height (Test no.
21) 368
Figure 157: Concentration within individual layers 369
Figure 158: Cumulative volume o f individual layers 369
Figure 159: Effect o f released mass on V V 370
Figure 160: Effect o f room area on V Y 370
Figure 161: Effect o f room air speed and discharge angle on W 371
Figure 162: Effect o f release time and post-ventilation period on V V 371
Figure 163: Effect of release height and air discharge height on VFtF 373
Figure 164: Comparison of measured and calculated maximum overpressures 380
Figure 165: Deviation between measured and calculated maximum overpressures 380
Figure 166: Examples of room overpressure development 381
Figure 167: Effect o f release mass with constant room size or constant C r™on consequence
382
Figure 168: Effect o f room area with constant release mass or constant C r™on
consequence 382
Figure 169: Effect o f room airspeed on consequence 383
Figure 170: Effect o f release height and airspeed on consequence 383
Figure 171: Effect o f various venting routes on reduction o f overpressure 384
Figure 172: Overall modelling sequence for the combined QRA 387
Figure 173: f]eak, P Fso i, P Fv and f* for different leak sizes at Ainfn = 0.4 h'1 390
18
Figure 74: Variation of fh* with room level for different leak sizes for off-cycle 391
Figure 75: f* for different leak times for a range of infiltration rates 391
Figure 76: Overpressure and thermal intensity for different leak times 393
Figure 77: Integrated consequence-ffequencies for different leak times 393
Figure 78: Overall ignition frequencies 394
Figure 79: Overall risk factor 394
Figure 80: Relationship between and total V V (r2 = 0.82) 405
Figure 81: Relationship between Cfmax and the proportion o f V V above 1 m 405
Figure 82: Correlation between PF* and V V (r2 = 1) 406
Figure 83: Relationship between PF* and Cfmax (r2 = 0.82) 406
Figure 84: Relationship between mean flammable mass and consequence (r2 = 0.99) 407
Figure 85: Correlation between Cfmax and consequence (r2 = 0.83) 407
Figure 86: Room orientation and reference directions 424
19
GENERAL INTRODUCTION
Background
In 1987 the Montreal Protocol on Substances that Deplete the Ozone Layer was ratified by the
majority of industrialised and developing countries, resulting in legislation to phase out their use
and production. This initiated a search for suitable replacements for chlorofluorocarbons (CFCs)
and hydrofluorochlorocarbons (HCFCs), many of which were used as refrigerants. Since then a
number o f fluids have become accepted as alternative refrigerants, specifically
hydrofluorocarbons (HFCs), hydrocarbons (HCs), carbon dioxide, and various others to a lesser
extent. These have been adopted both as single component refrigerants, and as multi-component
blends. Physical and chemical properties o f these fluids differ from those of the traditional CFC
and HCFC refrigerants, and since the design and performance o f refrigerating systems is affected
by these properties it is necessary to establish means of handling the resulting implications.
Additional considerations for system design and performance aspects have arisen due to other
legislation relating to environmental and safety matters. The work presented in this study
comprise the development o f methods that may be employed for assessing certain environmental,
performance and safety aspects implicated in the application o f these alternative refrigerants.
Motivation and description of work

The author w as employed by a producer o f alternative refrigerants, and in the course o f earning
out technical duties relating to the implementation o f these refrigerants, a number of issues
affecting the selection and application of fluids were found to be inadequate or open to misuse. A
general absence of efforts to develop suitable methods for correctly handling some of these issues
was evident, and this provided the intermittent motivation for the author to produce the present
study. The three issues dealt with are evaluation o f the global warming impact of refrigerating
systems, balancing o f refrigerating system components that use mixture refrigerants with
temperature glide, and investigating the flammability risk associated with use of HC refrigerants.
Whilst broadly related, the three subjects handled in this study are diverse and this constituted
three separate, self contained parts. As such, each part comprises an independent introduction,
literature review, main body o f work, conclusion and references, and the content is summarised
below.
Extending Methods fo r Evaluating the Global Warming Impact o f Refrigerating Systems.

An approach commonly termed, Total Equivalent Warming Impact (TEWI), is employed to
evaluate the contribution o f a systems global warming impact resulting from greenhouse gas
20
emissions from energy use and refrigerant release. The literature shows that whilst existing
procedures for evaluating TEWI vary, it is always vague in terms o f scope, assumptions and
detail. In producing a thorough and more complete approach to estimating emissions, a detailed
model was developed to include all possible sources of emissions throughout the equipment
lifetime. To this end, influencing parameters were documented, and sub-models developed to
accurately describe them. Using the model, the influence of a variety o f equipment design
features, external conditions, ageing, and repair and maintenance practices were examined in
respect to their effect on the global warming contribution o f refrigerating systems. Also, the use
o f the Global Warming Potential (GWP) quantity for TEWI calculations is known to have
shortfalls whereby it does not suitably represent the relative impact o f emissions o f different
substances. Therefore, the interpretation of the impact o f global warming gases was addressed
and alternative measures were adopted in place o f GWP.
Developing a New System Balancing M ethod fo r M ixture Refrigerants.

Conventional methods for testing, rating, data presentation and selection o f the main system
components (compressor, expansion device, condenser and evaporator) are currently based on a
standardised industry practice originally developed for pure refrigerants. System balancing
techniques, which utilise performance data, are based on the same premise. Currently, the
majority’ of commercialised alternatives refrigerants are zeotropic mixtures that display a
temperature glide during phase-change. This impacts on the performance rating o f refrigeration
components and as such invalidates the current methods listed previously, particularly for the
condenser and evaporator. The consequence of this is introduction o f errors in the design o f the
refrigeration system. Modified methods are proposed to be employed within the testing, rating,
data presentation and selection o f components, and these changes are integrated into a revised
balancing technique so that the behavioural aspects associated with zeotropic refrigerants, and
other influences such as pressure loss and superheat, are accounted for.
Appraising the Flammability Hazards o f Hydrocarbon Refrigerants.

HC refrigerants present fire and explosion hazards due to their flammability. To investigate this,
a quantitative risk assessment (QRA) model was developed to evaluate the potential for ignition
and consequences when hydrocarbons are employed in stationary refrigeration and air-
conditioning equipment. Initially, an approach for calculating ignition frequencies, leakage
frequency, probability o f flammable concentrations, and associated consequences, being
overpressures and thermal radiation was documented. Empirical data for leakage, hardware
failure and sources of ignition w as provided, as well as data from dispersion experiments was
21
used to develop sub-models. The findings from the dispersion experiments also allowed a method
to be developed for equipment design and installation features to assist with dispersing leaked
refrigerant. Various data were used to demonstrate the application o f the QRA model, and in
particular, to examine the effects that design, installation o f equipment and external conditions
have on the frequency of ignition and the associated consequences.
22
PART 1: METHODS FOR EVALUATING THE GLOBAL WARMING IMPACT OF
REFRIGERATING SYSTEMS
23
NOMENCLATURE
Letters
a constant
A area (m2)
AGTP absolute global temperature potential (K)
AGWP absolute global warming potential (W m'2 kg'1 y)
A area of HTF-side exchanger surface (m2)
HTF
area for two-phase heat transfer (m2)
b constant rating (W)
B app application basic rating (W K '1)
C constant
C global surface heat capacity (J K'1W '2)
COP coefficient o f performance (-)
COPcomp compressor coefficient o f performance (-)
Cp specific heat at constant pressure (kJ kg'1 K '1)
D diameter (m)
E energy consumption (kW h)
p anctl energy consumption o f ancillary components (kW h)
E comp compressor energy consumption (kW h)
F radiative forcing (W m'2 kg'1)

GDI global damage index (-)
GTP global temperature potential (kg C 0 2)
GWP global warming potential (kg C 0 2)
h heat transfer coefficient (W m'2 K '1)
H specific enthalpy (J kg'1)
specific enthalpy at evaporator outlet/compressor inlet (J kg'1)
specific enthalpy at evaporator inlet (J kg"1)
k thermal conductivity (W m '1 K '1)

K construction emission factor (-)
L length (m)
B trav travel distance (km)

M mass (kg)
M' emission mass flux of non-C02 gas (kg v 1)
M ” emission mass flux of C 0 2 (kg y '1)
M m total C 0 2 equivalent emission mass flux (kg y'1)
fnr refrigerant mass flow rate (kg s'1)
n index, sensitivity factor

N number
Non, number of components (-)
N xn. number of service/maintenance visits (-)
N start number of compressor starts (-)
N tpup number of refrigerant top-ups (-)
P pressure (Pa)
Ap pressure drop (Pa)
Pdisch pressure at compressor discharge (Pa)
Psuet pressure at compressor suction (Pa)
Pvent probability that refrigerant is vented (-)
O heat flow (W)
Q<m application heat load (W)
Qc condensing capacity (W)
Qc,comp compressor condensing capacity (W)
Qcomp compressor capacity (W)
Qe evaporating capacity (W)
Qc,comp compressor evaporating capacity (W)
AR mean variation in leak ratio (-)
Rdist leak mass fraction during refrigerant distribution (-)
Re Reynolds number (-)

D leak mass fraction during equipment operation (-)
oper
Sou solubility of refrigerant in oil (kg kg'1)
Sw wall shear stress (N m'2)
25
t time (s, y)
T temperature (°C)
At' reference time (s, y)
r variable temperature (°C)
Tamb ambient temperature (°C)
TOPP application temperature (°C)
Tdisch temperature at compressor discharge (°C)
TLT1 total lifetime temperature impact (K)
^ a fc off-cvcle time (s)
& onc on-cycle time (s)
*op operating time for compressor wear (s)
ATm thermostat differential temperature setting (K)
compressor start-up time (s)
T~* temperature at compressor suction (°C)
K wall temperature (°C)
u velocity (m s'1)
U overall heat transfer coefficient (W m'2 K '1)
V volume (m3)
internal or representative volume (m3)
vhose refrigerant hose/manifold internal volume (m3)
V
y HTF HTF volume flow rate (m3 s'1)
K swept volume flow rate (m3 s'1)
w power (electrical) (W)

w
” anal power o f ancillary component (W)
wcomp compressor power (W)
w
rr HTF pow er applied by HTF flow (W)
X heat exchanger wall/fouling thickness (m)

X mass fraction of component (-)
Y vear
26
Greek and other symbols
X component leak factor ( y 1)
e materials emissions factor (kgC 02 kg'1)
£ cpnt component manufacture emission factor (kgC 02 kg'1)
£ r refrigerant manufacture emissions factor (kgC02 kg'1)
^ bra: brazing/welding emissions factor (kgC02 kg'1)
E chemical reaction energy (J mol'1)
9 reduced charge fraction (-)
Îcpnt component (fan or pump motor) efficiency (-)
^lincin incineration efficiency (-)
volumetric efficiency (-)
A friction factor (-)
M viscosity7(Pa s)
car molecular mass of air (kg mol'1)
^ C 02 molecular mass of carbon dioxide (kg mol'1)
M *, molecular mass of oil (kg mol'1)
Mr molecular mass of refrigerant (kg mol'1)
VY mean deviation in annual emission factor (kgC 02 kW '1h '1)
VD mean deviation in daily emission factor (kgC 02 kW '1 h'1)
e, coordinate for position o f annual emission factor (°)
p density (kg m'3)
a climate sensitivity7parameter (-)
S transport emissions factor (kgC 02 km'1)
T atmospheric lifetime (y)

T time constant (y)
4 energy emissions factor (kgC 02 kW '1 h'1)
V proportion o f refrigerant emissions from refrigerant production (-)
Subscripts
app application
braz from brazing, welding, etc
chrg from refrigerant charging
cnst from equipment construction

cpnt relating to components
des design (rated)

disp from disposal
dist from refrigerant distribution

dmtl from dismantling
EOL end o f life
evac of refrigerant evacuated
foul fouling
H TF heat transfer fluid

irtcirt o f material from incineration
m anf from manufacturing
max maximum
min minimum
oil o f compressor oil
oper from equipment in operation
prod from production
ptst from pressure/leak testing
r refrigerant
recov o f recovered refrigerant
s saturated
serv from service and maintenance activities
SOL start of life
sys system
Tot total
trav from travel
Use in-use
vent of vented refrigerant
x species
SECTION I: INTRODUCTION
1.1 Environmental burden of refrigerants

With the discovery of the link between CFCs and HCFCs and the depletion o f the ozone layer, the
United Nations Environment Programme (UNEP) formulated the Montreal Protocol in 1987, to
phase-out the use and production o f these substances. In response to the Montreal Protocol,
alternative refrigerants were sought, and this search produced a number of potential substances
for applications where only CFCs and HCFCs w ere previously used: HFCs, HCs, ammonia and
carbon dioxide (CO:). Simultaneously, attention turned to the issue o f climate change, and in
particular, global warming. Subsequently, the Kyoto Protocol was developed under the UN in
1997, which prescribes the limitation and reduction of emissions of a group o f anthropogenic
“greenhouse gases"’ (GHGs): C 0 2. nitrous oxide (N20 ), methane (CH*), HFCs, perfluorocarbons
(PFCs) and sulphur hexafluoride (SF6). Whilst the Kyoto Protocol has not been ratified to date,
numerous governments have already published national legislation to meet the Kyoto targets for
emissions reduction. In order to quantify the contribution of these gases to climate change, the
initial international discussions on climate change in 1990 adopted the use o f Global Wanning
Potential (GWP) of the gas (IPCC, 1990). GWP is a measure o f the insulating properties that a
gas has on the heat radiating away from the surface of the earth, and is relative to the effect
caused by one kilogram o f C 0 2.
In response to these political developments, GHG emissions associated with refrigeration systems
need to be assessed, particularly for comparing and screening alternative refrigerants and
systems. Given that conventional vapour compression systems could be responsible for emissions
of C 0 2 (from power stations) and also leakage o f refrigerant (such as HFC's), the global
w arming effects o f each gas could be added together to evaluate the overall impact o f
alternatives. Such an environmental assessment technique has developed into a method commonly
known as ‘Total Equivalent Warming Impact” (TEWI), although slight modifications to the
method have resulted in other terms, including ‘‘Life Cycle Warming Impact” (LCWI) and “Life
Cycle Climate Performance” (LCCP).1 In its most basic form the calculation can be expressed as
equation (1).
= ■GWPt ) + M ” (1)
1Hereafter, the TEWI term will also refer to these other variations, except otherwise stated.
29
where M is the mass of substance emitted; M m represents the total emissions in terms of CO 2
equivalent, for each o f the refrigerant species i , M ' is the amount leaked, and M ” is the total
mass o f CO: emitted, t\picallv due to energy consumption by the system.
To date, TEWI is used frequently by industry' and Government as a tool for comparing the
‘‘strength'’ o f GHG emissions o f systems and equipment.2 There is increasing reliance on TEWI,
despite many omissions evident in the procedures that are currently' employed. Several
researchers have questioned the suitability7and value of TEWI (e.g., Kruse, 1999) but from
surveying the literature, it is known that little attempt has been made to resolve many of the
issues. Given the escalating importance o f estimation o f GHG emissions, there is a clear need for
improving the current method.
1.2 Literature review

A literature review was carried out to provide a general overview o f the use and application of
TEWI and associated methods. In particular, it was used to identify the level of detail and
complexity adopted for calculation, means o f handling certain variables, sources o f input data,
etc. Additionally, studies on practical uses of TEWI were considered to indicate the context
within which TEWI is normally used, and how7the results o f comparative assessments are
normally handled. Lastly, it w'as deemed important to understand the basis for quantification o f
the global w arming impact o f the gases involved, and any limitations that could affect the
interpretation o f a TEWI assessment.
1.2.1 Existing methodologies for evaluating environmental impacts

Following the development of the GWP to help compare environmental impact of CFC
alternatives (see Section 1.2.3), the TEWI concept was an obvious progression and the evaluation
process for alternative refrigerants began to include indirect GHG emissions from energy7
production required to operate the system. The actual numerical methodology7was formalised by7
a UK industry7publication (BRA, 1996), and later within a European Standard on safety7and
environmental requirements for refrigerating systems (EN 378, 2000). The BRA guidelines
provided some background information, and a series o f numerical procedures and supporting
data, to estimate refrigerant leakage, system energy consumption and corresponding CO 2
emission rates. Nevertheless, the guidelines were considered to be rudimentary and lacking the
2 Requests for TEWI comparisons have become part of the tendering process for RAC equipment, and as
the basis for cost-effectiveness studies commissioned by UK Government departments.
30
consideration o f many influencing parameters (Colboume and Suen, 1999).3 Awareness of these
peripheral parameters was further highlighted in the UNEP report on alternative refrigerants for
ozone depleting substances (ODS) (UNEP, 1999).
Papasaw a and Moomaw (1998) proposed a method that extended the conventional TEWI
approach, and named it ‘“Life cycle warming impact” (LCWI), which also included emissions
associated with refrigerant production, transportation o f materials, and refrigerant recycling.
They also included for the breakdown o f refrigerants in terms of the GWP of decomposition
substances; this indirect contribution is now included in the more recent GWP values
(IPCC/TEAP, 2005). The description provided by Papasaw a and Moomaw was comprehensive,
although repeatable use o f the method would be difficult due to absence o f key calculation
procedures and input data. O f importance was an example study o f domestic refrigerators, which
showed that some o f the parameters normally ignored did have a significant impact on the result.
Conversely, Aprea et al (1999) present a simplified criterion to determine TEWI that used
graphical methods.
More recently, Heikkila and Fahlen (2003) and Bamabe et al (2003) have proposed life-cycle
assessment (LCA) methods to help estimate the emissions associated with the equipment
throughout its life. This was used to assist with optimising the design to minimise the
environmental impact o f the equipment at design stage. Instead o f limiting its scope to GHGs,
these studies consider other impacts normally included in an LCA (e.g., acidification, volatile
organic compounds, etc.), and each of these is subject to weighting to produce a single “overall
environmental impact” value. Again, in both studies, the concept is described comprehensively
with examples, but data and procedure were largely omitted. Another method proposed by Bodyal
and Dubro (2003) used this parameter-weighted approach and included additional measures such
as system “efficiency” and safety aspects such as flammability and toxicity. The weighting
assigned to each parameter was determined subjectively. The concept and general method of LCA
is covered by a series o f international standards (e.g., ISO 14040), from which a procedure for a
specific application can be developed.
One other variation was a method described by Zhang and Reistad (1998), and named “Total
equivalent resource exergv” (TERE). This takes the sum of exergy production for the
manufacture and operation o f a system, but also includes the exergy production due to the
recovery of emitted CO 2 (based on the GWPs o f the various GHGs). With this additional CO 2
3 A paper published by the present author/tutor resulting from the initial study for this thesis.
31
recovery element, the method could be used to evaluate the efficiency of a system in its use o f
natural resources (from a global wanning standpoint). Worked examples for some applications
were used to demonstrate the difference between the results of TEWI and TERE, the most
notable o f which was that the influence o f manufacture of equipment was much greater using
TERE
1.2.2 Survey of comparative studies

The majority o f the literature citing TEWI is normally concerned with performance comparison
o f different refrigerants and/or systems. Large volumes describe experimental and/or theoretical
studies on system energy consumption, which are complimented by presenting a short TEWI
comparison. Such exercises normally adopt a simplistic approach where energy consumption is
converted to equivalent C 0 2 emissions for some operating period, and rarely provide a useful
contribution to understanding o f the method itself. As such, only those studies that cover
additional considerations in terms of TEWI evaluation were review ed presently.
Perhaps the most widely promoted TEWI studies are a series of comparisons commissioned by
the Alternative Fluorocarbons Environmental Acceptability Study (AFEAS), initially in 1991
(Fischer et al), then in 1993 (Fischer et al) and again in 1997 (Sand et al). These were reports
concerning several sectors including domestic and commercial refrigeration, car and residential
air conditioning, and chillers. A latest report (Sand et al, 2002) repeats the results of the previous
studies. Evaluations were carried out to compare the impact of various different refrigerants for a
range o f different geographical locations, using corresponding local conditions (temperatures,
designs, power station emissions, etc.) as inputs to the calculations. Despite the broad range of
scenarios, the description of the actual methodology, source data, assumptions and calculations
was very limited. Therefore, whilst large amounts of output data w ere reported, it is questionable
as to how useful reports such as these are.
Use o f TEWI for comparative purposes is demonstrated in a study by7Hwang et al (2004) which
compares a low GWP refrigerant (R290) and two HFCs (R404A and R410A). Experiments were
conducted to measure the performance in a particular type o f system, which were optimised for
each refrigerant. The test data showed the COP for R290 w as more than 10% above that of the
HFCs. However, a 10% cost deficit was assumed for the system using R290 due to flammability
mitigation, and this was offset by giving the systems using HFCs a larger condenser. The
performance data was re-modelled (causing the HFCs to have superior COP under most
conditions), and using an unusually low7leakage rate, concluded that the TEWI for R410A was
32
lower than R290. Since the size o f the cost deficit is known by this author to be overestimated
(Colboume, 1999), and that the annual leak is unrealistically low for the equipment used, the
emphasis on a 4% lower TEWI is unjustified given the uncertainties associated with the entire
TEWI calculation.
A more constructive use o f TEWI has been reported by Benstead (1996), where the TEWI for
two different supermarket systems were calculated and then independently audited. The
retrospective auditing process was useful since it found that initial assumptions about fan power
were underestimated and that their assumption o f an average ambient temperature leads to
overestimation. Another use of TEWI applied to supermarket installations was as a design tool
(Davies, 2004). Various system designs w ere evaluated from a TEWI perspective in order to
determine the design that would lead to the least GHG emissions. A type o f indirect system using
low-GWP refrigerants was found to achieve 26% lower TEWI than the conventional direct
expansion approach.
Zhelenzv at al (1996) evaluate different refrigerants for domestic refrigerators, but pay attention
to causes of efficiency differences such as component pressure losses. Another use o f TEWI was
by Aprea and Greco (1998), who carried out thorough tests o f a particular system using different
refrigerants to enable accurate estimation o f the energy contribution. Based on fixed parameters
such as leak rate, TEWI for R22 was evaluated, and by working the method backwards, the
required minimum system efficiency for R407C was determined if it w ere to equal the R22
system's TEWI. The results also showed how the difference in TEWI for R22 and R407C
systems varies with ambient conditions and operating mode (i.e., cooling or heating). A
computational approach was reported by Gopalnaravanan and Rolotti (1999), where eight
different refrigerants were compared using a simulation model to predict energy consumption. A
fixed system design was used for all refrigerants, and only compressor swept volume was
adjusted to maintain equal refrigerating capacity at rating conditions.
Some other studies have used an extended LCA approach where additional environmental
impacts were addressed. Hansen and Gustafsson (1994) and Cintar and Hadfield (2000) both
focussed on the manufacture and operation of the compressor in small hermetic systems. The
former addressed the manufacture and disposal o f several compressor models, which showed
similar results for each, not just for GHG emissions, but also for sulphur emissions, water usage
and others. It was also shown that by sacrificing a greater impact at design stage (i.e., use of
more materials), this could dramatically offset emissions during in-use stage (i.e., less energy
33
consumption), whilst the scrapping phase had negligible impact. Cintar and Hadfield considered
many environmental impacts in comparing compressors for HFC-134a and HC-600a. The
analysis showed significant reduction in GHG emissions when using HFC-600a, which was
attributed to different compressor design and lubricant types, as well as the refrigerant itself.
Johnson (1998) evaluated HFCs and HCs in car air conditioning to compare TEWI and emissions
of volatile organic compounds (VOCs). The analysis used a comparatively thorough “cradle-to-
grave” approach, and found that the HC resulted in significantly lower GHG emissions,
particularly when refrigerant leak rates were high, and VOC emission were found to be similar. A
more comprehensive study was a LCA by Frischknecht (1999) that covered heat pumps, building
air conditioning, industrial deep freeze and supermarket refrigeration applications, with particular
focus on comparing “natural” (HC, ammonia) and synthetic (HCFC, HFC) refrigerants. Actual
products/installations were analysed in terms o f individual components and their routes to the end
product. A LCA database was used to determine all emissions associated with each component,
and other emissions resulting from the system operation. It wras found that energy' consumption
was the dominant factor for most emissions (>50%), and therefore system efficiency was an
important consideration. In general, systems using HC and ammonia resulted in less overall
emissions when compared to fluorinated refrigerant systems. Again, effort focussed on presenting
comprehensive results rather than detailing methodology that prevented a full appreciation o f the
results.
1.2.3 Measures of climatic impacts

The emission o f anthropogenic GHGs results in changes in global surface temperatures, which
ultimately leads to climate change, or “an average change in the earth's weather” (IPCC, 2001).
Understanding the physical processes that occur in the atmosphere, on land, and in the oceans is
the subject o f massive research.4 In assessing the impact of emissions, it is necessary to have a
measure that relates climate impact to a given mass o f a particular species, for example, the
impact of a 1 kg emission of HFC-134a. As mentioned in the introduction, the internationally
agreed measure was GWP, and this has been employed in impact assessments since. However, as
implied by its definition (IPCC/TEAP, 2005), GWP is essentially a measure o f the property7of a
species under a given set o f conditions, and thus a precursor to the actual climate impact:
4Most of the important research is collated and assessed by the Inter-governmental Panel on Climate
Change (IPCC). who publishes report everv four years, the latest being the Third Assessment Report
(IPCC, 2001a).
34
“An index, describing the radiative characteristics o f well-mixed greenhouse gases, that
represents the combined effects of the differing times these gases remain in the
atmosphere and their relative effectiveness in absorbing outgoing infrared radiation. This
index approximates the time-integrated warming effect o f a unit mass of a given
greenhouse gas in today’s atmosphere, relative to carbon dioxide.”
Since GWP was adopted, its suitability to represent the actual impact of the release have been
questioned, and extensive research since 1990 has been unable to draw a conclusion on whether
any other method is more suitable than GWP (O’Neill, 2000). This is mainly because it is unclear
how effectiveness o f a measure should be judged. For example, a large number of perceivable
physical impacts exist, such as a change in surface (air) temperatures, weather behaviour, or
economies, etc., where the size o f the change o f each impact varies between gases. A further
consideration is the time scale o f the impact: the rate and the duration of the change, specifically
since the impact of an emission may be more sensitive to the rate o f surface temperature change,
than the absolute change (Godal, 2003). For example, O’Neill notes that the use of GWP to
compare two gases (of equivalent emissions) results in different climate responses, i.e.,
disproportionate rates and magnitudes of change across the different types of impacts. In fact,
emissions o f two different gases with the same equivalent GWP did not produce the same climate
impact response.
In terms of the formulation of GWP, various problems exist with its quantification. First, most
common manmade GHGs break down in the atmosphere and therefore have a finite atmospheric
lifetime ( r ), although it varies according to the atmospheric concentration o f itself and other
gases. CCK on the other hand remains chemically stable under normal atmospheric conditions,
and therefore has an infinite r . Instead of chemical decay within the atmosphere, removal of CO 2
is reliant upon “carbon sinks” such as the ocean and plants that absorb it from the atmosphere,
resulting in an apparent r ~ 500 years. Reference to Table 1 shows that for common refrigerants
r are less than 50 years. As implied by the definition above, GWP is the ratio of a species’
absolute GWP (AGWP) to the AGWP o f CO 2 , and AGWP is determined by quantification of the
radiative forcing of the gas over some reference, or integration time (t).
For gases with a finite r , calculation o f AGWP using a t > r tends to produce no additional
contribution from when t = r , whereas AGWP for a gas with infinite r (i.e., C 0 2) always
increases with a longer t . Thus for a given species, GWP( t ) tends to reduce as it is evaluated for
t that exceeds the species’ r , since its AGWP thereon remains approximately constant whilst the
35
AGWP o f C 0 2 continues to rise. This dependency o f interpretation of a gases global warming
impact on the largely subjective choice of integration time means that the results of calculations
that use GWP are implicitly subjective themselves. Another parameter that introduces uncertainty
into GWP is the basis for calculation of radiative forcing ( F ) which is dependent on atmospheric
concentrations o f other gases. F for a given species is calculated using climate models, but these
models were based on an atmosphere that is composed o f gases at present day concentrations.
Since concentrations o f all GHGs will be different at some time in the future, values o f F - and
therefore GWP - will be different (Smith, 2003). According to a review of GWPs within
IPCC/TEAP (2005), the net uncertainty could be as high as ±50%, with longer r gases tending
towards smaller uncertainties.
Table 1: GHG (refrigerant) data from IPCC TEAP (2005)
Molecular Radiative forcing, fr Atmospheric GWP(100)

Name Mass lifetime, x
(W m'2 ppb'1) (W m'2 kg'1) (kgC02 kg'1)
(kg kmol"1) (year)
R12 120.9 0.320 1.49 x 10'11 100 10720
R22 86.5 0.200 1.30 x 10'11 12.0 1780
R32 52.0 0.110 9.73 x 10'12 4.9 670
R124 136.5 0.220 9.06 x 1012 6.8 600
R125 120.0 0.230 1.08 x 10'11 29.0 3450
R134a 102.0 0.160 8.27 x 10'12 14.3 1410
R143a 84.0 0.130 8.70 x 10'12 52.0 4400
R152a 66.1 0.090 7.66 x 10'12 1.4 120
R227ea 170.0 0.260 3.31 x 10'11 34.2 3140
Ethane R170 30.1 0.0032 5.98 x 10'13 0.214 8.4
Propane R290 44.1 0.0031 3.95 x 10'13 0.041 6.3
Iso-butane R600a 58.1 0.0047 4.55 x 10'13 0.019 7
Propylene R1270 42.1 0.034 4.53 x 10'12 0.001 4.9
Carbon dioxide 44.0 1.54 x 10'5 1.98 x 10'15 See text 1.2.3 1
Ammonia R717 19.0 0 0 0.008 0
7 To convert from W m'2 ppb'1 to W m'2 kg'1, multiply by /(5.15 • 109 • M x)
In order to overcome some of these limitations several variations o f GWP have been proposed.
Lashof and Ahuja (1990) suggest using an infinite lifetime integration period, and conversely,
Hammond et al (1990) consider the use of one-year integration period appropriate since it
accounts for the instantaneous warming effect. The second assessment report o f the IPCC (1995)
36
considers the use of absolute GWP since this has the advantage that the impacts attributed to
non-CC>2 gases were less dependent of the accuracy of the CO 2 evaluation. Wigley (1998) goes
back a stage and proposes a Forcing Equivalent Index (FEI), which is based on the rate o f change
o f F required for achieving a specified increase in temperature.
In addition to the parameters closely related to the properties o f the GHG (e.g., r , F and GWP),
there are also other indices addressing physical impacts (such as change in sea level) or economic
impacts. A new approach to estimate the global temperature change potential (GTP) due to an
emission of a GHG is presented by Shine et al (2004). In their work, the radiative forcing effects
o f a GHG were related to the available thermal mass o f the earth's surface to calculate
temperature rise. Taking on the effects of temperature rise. Smith (2003) states that the most
important approach to evaluating the impact of GHGs is economic efficiency, suggesting that the
cost o f impacts is related to the square o f the temperature rise integrated over a given timescale.
Fuglestvedt et al (2003) also describes this approach, and notes that the index could also include
abatement costs so that an optimum or minimum cost for emission reduction could be determined.
They refer to several authors who evaluate the damage impact based on an exponent of
temperature increase, to a value between 1 and 2. This could be a static (fixed time period) or
dynamic approach, so that the time-dependent or accumulated damage costs can be calculated.
Reilly et al (1999) tried to combine scientific and economic aspects in proposing a variable
economic index which accounts for particular emission scenarios and costs arising from
abatement policies and physical damage caused by climate change. However, the authors point
out that the use of such an approach requires a link to the global economy, such as two-way
adjustments with interest rates. In evaluating a measure termed “Global Damage Potential”
(GDP), Frankhauser (1994) found that for a given GHG, the GDP typically equated to GWP.
Conversely, Reilly and Richards (1993) found that the contribution o f non-C02 gases to damage
increased with global temperature rise compared to CO 2 , since C 0 2 promoted (beneficial) plant
growth, thus increasing the size o f the carbon sink. Other variations propose an index where
abatement costs were evaluated by ensuring a given impact parameter does not exceed a given
value. For example, Manne and Richels (2001) estimate a cost index to ensure a given
temperature change is not exceeded by a particular year.
1.3 Formulation of objectives

There is a general acceptance of TEWI by RAC equipment manufacturers, installers and end-
users wanting to evaluate the environmental impact of systems, and subsequent choices and
37
designs are influenced by the results. The literature revealed that current methodologies
necessitate broad assumptions and interpretations of subsequent results tend to be shallow. Due
to the often perceived significance of TEWI calculations suggests that the concept demands
deeper consideration. The literature review indicated a number of key areas associated with
TEWI calculations that are not normally addressed in detail and therefore it is necessary’ to
develop an approach to resolve these issues.
1.3.1 Summary of findings

There are several key findings drawn from the literature review, as follows.
- Existing methodologies. The conventional TEWI concept and procedure, such as that
detailed by BRA (1996), is found to be limited in several respects and omit a large number of
influencing factors. Whilst the basic formulation is established, written methodologies are
scarce and when presented, input assumptions are vague making them open to broad
interpretation. This is evident when the vast number of TEWI evaluations found within the
published literature that generally have the objective of comparing alternative refrigerants but
frequently report conflicting conclusions.
- Comparative studies. Few studies detail specific input values, or indeed how7they are
handled; there w ere no models described in sufficient detail to enable reported results to be
reproduced. Also, many assumptions adopted by the various studies are often without
justification, and is particularly the case for choice of annual leakage rates, which for similar
systems vary widely and appear to be chosen arbitrarily. Many peripheral parameters
(outside the usual system leakage and energy consumption) are omitted on the basis that they
are “negligible'’, although this is rarely justified. Further, when very small differences in
comparative results o f TEWI calculations (e.g., < 5%) are obtained, they are used to defend
the selection of a technology that is found to give the low est impact.
- Climate impacts. The validity of drawing conclusions on this basis is further questioned in
light o f the uncertainties and subjectivity associated with the determination o f GWP, and
indeed, its suitability o f accurately representing the climate impact of the various gases under
consideration. This point is particularly relevant because of the different characteristics of
common refrigerants wiiich are short-lived with high radiative forcing, and C 0 2 with virtually
infinite lifetime and low radiative forcing. To address this, a number o f alternative measures
38
o f global warming impact were identified that can be adopted as an alternative where they
overcome some o f the drawbacks associated with GWP.
The findings from the literature review indicate an absence o f an otherwise formal approach to
calculating TEWI, in terms of level of detail and selection o f influencing parameters. Similarly,
comparative results are interpreted in a largely subjective manner, due to neglect of certain
influencing factors resulting from misconceived assumptions. The current work is intended to
address these aspects by identifying a broad range of influencing parameters and assessing the
validity o f their inclusion in TEWI calculations, and further, to investigate alternative means of
interpreting the global warming impact.
1.3.2 Aims and objectives

The purpose of a TEWI calculation is to evaluate the contribution of an RAC system to global
warming, and thereby help identify a suitable refrigerant and/or system that provide the lowest
impact. It is not considered necessary to develop a different concept since the current
fundamentals of quantifying GHG emissions from individual sources and summing them is
sound. Instead, the objective is to seek improvements to better represent the impact of an RAC
system.5 This was achieved by a thorough assessment of factors that could influence GHG
emissions, where their relative contributions are evaluated by compiling them within a
mathematical model and comparing the outputs under a range of conditions. The model also
includes alternatives to GWP so different measures o f global warming impact can also be
considered.
Selection of influencing parameters

Emissions are currently assumed to result solely from the operation of a refrigerating system,
which neglects other stages and activities present during the life of a refrigerating system. Other
stages in the life o f the equipment (such as manufacture) will also involve energy use and gas
emissions, and will therefore contribute to the overall impact to global warming. There are further
peripheral sources o f emissions that occur solely because o f the equipments’ existence. Thus, the
boundaries o f the assessment should be broadened to consider all emissions for which the
refrigerating system is ‘‘responsible”. All the parameters should be clearly identified, which was
achieved by working through o f each process and activity associated with the life o f the
equipment.
5 The case of automotive air conditioning and transport refrigeration is not considered here.
39
Development of emissions model
Calculations are currently made using limited input data, so the outputs must be considered
general, rather than accounting for the design features o f specific equipment. The usefulness of an
extended methodology can be enhanced if it accounts for controllable design and operating
parameters. This can be achieved by quantifying the additional (peripheral) parameters that
influence emissions as discussed. Where a specific parameter is considered to make a significant
contribution to the net result, it should be included in the model. It is important to have a basis for
determining what a “significant” contribution is, and to clearly state and justify all assumptions
and include uncertainty analysis over a range o f values about the assumed input value.
Subsequently, the model can be used to answer questions about factors that influence TEWI,
such as how it is affected by design, operating and materials handling parameters.
Interpretation of climate change

The use of GWP has limited application because the GWP of most refrigerants is not necessarily
equivalent to the GWP of CCk Also the concept of GWP has little physical meaning (outside the
context o f climatology), so a more meaningful measure should be used, and its interpretation be
explained. In the first case, AGTP was chosen to represent global warming impact because it has
perceivable output (i.e., change in temperature), and it has a clear time element - both of which
help quantify consequences of global warming, such as by means o f a damage index. Thus, by
incorporating alternative measures of climate impact, the results o f the emissions model can be
reinterpreted in terms o f temperature rise and damage.
40
SECTION 2: DEVELOPMENT OF EMISSIONS MODEL
2.1 Modelling approach

The basic approach to determine TEWI is to estimate the emissions from system leakage and
energy consumption, as given in BRA (1996). However, many additional influences may
contribute to the total emissions other than those resulting directly from the system operation.
Different sources o f emissions are illustrated in Figure 1, where the life of the equipment is
broken down into three stages: start of life (e.g., production), in-use (e.g., operation, servicing),
and end of life (e.g., disposal). For these stages, each source that causes emissions is shown and
the type of emission is indicated, i.e., CO 2 from energy production and refrigerant releases. In the
model, total lifetime emissions o f refrigerant and CO; are the sum o f emissions for all the
elements, as equation (2) and (3) respectively.
(2)
(3)
where subscripts SOL , U se, and EOL represent the emissions from start of life, in-use and end
o f life stages for each element, i , that is a source of emissions.
A mass balance for CO; emissions is not necessary, but because any refrigerant produced will
either be emitted to the atmosphere or destroyed by incineration a mass balance o f refrigerant
throughout its life is needed. Equation (4) gives the total mass of refrigerant “consumed" over the
life o f the equipment; refrigerant recovered at EOL is not included because it is assumed to be
recycled for another purpose.
M r .Tot ~ M Tot incin (4)
where M incin is the mass o f refrigerant incinerated at EOL.
When evaluating the emissions from any of the possible sources in Figure 1, limiting boundaries
must be established in order to determine the accurate relationships between the source and the
emissions, whilst avoiding unnecessary detail. For example, emissions associated with the
manufacture o f a system component may be related to the mass o f the material used to make it,
such as from energy required by machinery and flue gases from heat production. However, to
enable a factory to operate, staff are required and the facilities where they work will produce
41
emissions, as would the manufacture of machinery used within the factory . Within the
boundaries, each source is characterised according to observations based on the authors and
others' experience relating to RAC equipment, and the availability o f data. Uncertainties are
accounted for by identify ing a likely range associated with each source, and a maximum and
minimum input value associated with the size o f the emissions employed. Thus, the output o f the
model will be a range o f maximum and minimum masses.
o w
o £
S' oq"
Start of life (SOL) t <§
>2 S-
*1 a
manufacture 1— 1 refrigerant 1— 1 production |— --- r*------ •
11------— — —| distribution |—- --- — ----
----------------- 1 component 1— 1 production |— —
1----------------- 1 assembly |— -----•
construction |— [ installation |— | brazing, etc ]— —
----------------- 1 refrigerant handling 1— 1 charging/leak test 1— — — ----- »
In-use (Use)
operation 1— i compressor/ancillary 1----------------------------------- -----m

1 ”1 i 1
I leakage |
servi cing 1— j refrigerant handling 1— 1 recovery 1—
r ■■ .......... ^ "i
----------------- 1 evacuation |—
-----------—— charging'leak test — ----:--------•
----------------- 1 replace components |— | brazing, etc |—
----------------- 1 production I— ---- •
----------------- 1 iransponation |----------------------------------- ---- • ;
End of life (EOL)
dismantling 1— 1 refrigerant handling 1— 1 recovery 1— — — ----- •

----------------- 1 evacuation |— — — ---- «
----------------- 1 venting |— ------ ------ •

----------------- disassembling |— | brazing, etc |— — •
----------------- 1 transportation |-----------------------------------
disposal |— j incineration |----------------------------------- — -m?----- «
Figure 1: Identification o f sources and types o f GHG emissions
42
2.2 Start of life stage
Start of life stage comprises production of raw materials, component manufacture, assembly,
installation and charging with refrigerant. The general process is detailed in Figure 2, where the
shaded boxes represent sources of emissions. Emissions of refrigerant and C 0 2 at SOL comprise
emissions from two general sub-stages: component manufacture ( M manf) and equipment
construction ( M cml). equation (2) and (3) respectively.
m 'sol = m :„, + m ;.„ (?)
+a c „ (6)
Su-.rt
Charge refrigerant'.’
‘ K ll fcV'fi'JIi'ii.Hi l i 'l l i t i n 'i i i l r t f i B j -
Is relfigerant leaking?
Charge refrigerant'’
W ttm S E S S a m S tK tm
1_______
'End
Figure 2: Activities at start o f life stage causing emissions
2.2.1 Refrigerant and component manufacture

Refrigerant manufacture emissions occur in the production process (M 'prod ) and during
distribution, and storage (M 'dist) (equation 7). C 0 2 emissions also occur in refrigerant production
43
(M ”prod) and therefore contributes to , although the main source is from the production o f
materials and forming of system components ( M 'pnl) (equation 8).
M 'manf = M 'prod + M 'dist (7)
(8)
Refrigerant production, storage and distribution
Manufacture of refrigerants causes emissions associated with energy requirements and leakage of
feedstock and other materials. Data have been found for various refrigerants, either as quantities
o f individual species emitted per unit mass o f a given refrigerant, or expressed as total mass of
CO 2 equivalent. This data is listed in Table 2, where values for all refrigerants are given in C 0 2
equivalent { e r ) using GWP(IOO) for the various species. Several sources had been found for
certain refrigerants so the data was summarised by giving a maximum, mean and minimum value.
Where only one value of e r was found in the literature, the minimum and maximum values were
based on the average variation found for R22 and R134a (i.e., 0.45 and 1.55 respectively), since
these were supported by a larger number o f sources. Where there was complete absence of data
for a refrigerant, values were estimated using the mean ratio o f R22 and R134a s r to their GWP,
i.e., £r = 0.12 x GWP(IOO) of the refrigerant concerned. This is justified because the
manufacturing emissions are largely influenced by plant leakage and therefore is strongly
dependent on refrigerant GWP (Campbell and McCulloch, 1998). These data were then used to
estimate additional emissions resulting from the total quantity o f refrigerant used throughout the
system life. To be consistent with the current approach of calculating refrigerant and C 0 2
emission separately, the production emissions were split between C 0 2 and other species, but for
simplicity these were grouped for the selected refrigerant only, so that refrigerant emissions are
converted using its GWP(IOO). Using the references that provided a breakdown of each species
emitted, an average split between C 0 2 and others w as found. In general the larger proportion of
GHG contribution is from non-C02 gases, and the proportion of refrigerant GWP(IOO)
contribution is, y/ =0.17. The mass of refrigerant (equivalent) emitted is therefore calculated
from equation (9) and C 0 2 from equation (10).
(9)
GWP( 100)
44
M 'r-prcd= ( \ - v ) - M rJ1>rs r (10)
where M rJot is the total refrigerant consumed (equation 4).
Table 2: Manufacturing emissions o f refrigerants
C 02 equivalent emissions per unit mass of refrigerant, e r (kg C02 kg'1)

Refrigerant
Value Source Min Mean Max
R12 No data - 553* 1229* 1904*
205.3 Frischknecht (1999)
R22 205 299 393
393 ARAP (2004)
R32 190 1 Frischknecht (1999) 86 * 190 295*
R124 No data - 321 72* 111*
OO
(Sj
R125 160 1 Frischknecht (1999) llx 160
Banks and Sharratt (1996)
38.8
McCulloch and Campbell
84.5
(1998)
R134a 72 10 49 87
Frischknecht (1999)
87
ARAP (2004)
10
McCulloch and Lindley (2003)
R143a 120 T Frischknecht (1999) 541 120 186*
R152a No data - 6* 14* 22*
R227ea 120 Banks et al (1998) 541 120 185*
R404A 136 Frischknecht (1999) 611 136 211 *
R407C 142 Frischknecht (1999) 64x 142 220*
R410A 173 Frischknecht (1999) 78* 173 268*
R170 No data - 0.9* 1.2* 1.5*
1.5 Gover et al (1996)
R290 0.9 1.2 1.5
1.5 McCulloch et al (1998)
R600a 0.9 1.2 1.5
1.5 Gover et al (1996)
2.0 McCulloch et al (1998)
R717 2.0 2.3 2.5
R744 (CO2 ) 1.62 Frischknecht (1999) 0.7* 1.6 2.5*
R1270 1.3 SAEFL (1998) 0.6* 1.3 2.0*
* Estimated from data for blends. * Estimated from R22 and R134a s r and/or GWP data
45
Refrigerant losses during storage and distribution are assumed to occur at SOL stage for the
entire refrigerant used throughout the equipment life: although some of the refrigerant is actually
used during In-use or EOL, their production is accounted for here. Emissions due to refrigerant
production are calculated from the emitted proportion ( Rast) of the entire mass of refrigerant
used throughout equipment lifetime (equation 11).
M'dlst= R dlst-M rJot (11)
Refrigerant emissions occur during storage at the production facility, transfer to containers for
transportation, factory storage vessels and from cylinder filling. An analysis o f these activities
has been made, and it is estimated that a loss o f R ast = 0.0035 occurs for refrigerant going into
cylinders, and R ^ = 0.052 for refrigerant from bulk storage (Ritter, 2004). The higher losses
from bulk storage are largely due to leakage from piping circuits at the equipment manufacturers'
facility. The value o f RJist is chosen based on the charging method for the equipment.
Table 3: Component construction data
Component
Parameter
Evap. Cond. Compr. Oil Valves Pipework
a 5159.5 5801.0 17958 112.61 142.91 8933.0
Values
for a (upper) 9188.5 10295 27861 193.70 410.36 8933.0
equatio a (lower) 2897.1 3268.9 11575 65.473 49.769 8933.0
n (12)
n 0.8326 0.8275 0.6336 0.4937 0.6554 1.0
Aluminium 0.1 0.1 0.02 0 0 0
Materia
1 mass Copper 0.4 0.4 0.08 0 0 1.0
fraction Steel 0.5 0.5 0.90 0 0 0
(X )
Brass 0 0 0 0 1.0 0
AIA, AIA. Bitzer. Bitzer. Danfoss. ASHRAE
Source
Searle Searle Danfoss, Danfoss. NRS (2002)
Component manufacture
Evaluation o f emissions from component production is a complex process and has therefore been
simplified by conducting the estimation by mass o f primary material. The actual raw materials
may be sourced from a number o f different countries, and be subject to local processing practices
that differ widely (Norgate and Rankin, 2000). Throughout mining, extraction, purification, etc.,
materials may be transported over a variety of distances. Given the large number of possibilities
prior to arrival at the component factory, aggregate emission figures for primary materials have
46
been used. Both C 0 2 and refrigerant emissions are associated with the manufacture of raw?
materials and production/assembly of actual system components, but are all converted into
kgC 02 equivalent.
1000 T 10000 T
A compressor o evaporators
1000 -
ou
jerlow er limits
for compressor
upperlower limits
for condensers
upperlower upperlower limits

limits for oil for evaporators
0.1 -I—
0.000001 0.0001 0.01 0.0001 0.01
Compressor cyinder vokme (m3) Iiternal wlume (m3)
Figure 3: Data fo r compressor and Figure 4 : Data fo r condenser and evaporator

compressor oil mass mass
For system components, data for production of primary materials is used, and the mass of a given
material is dependent upon the mass of the component. Product catalogues w ere used to obtain
the mass of compressors, evaporators, condensers, pipework and valves (and ancillary
components). The mass was correlated against a representative internal volume o f the component
in the form of equation (12), observed to best represent the data.
( 12)
where a and n are a constant and index, listed for the different components in Table 3.
Despite the range o f components from different manufacturers, there appears to be a relatively
good correlation with the chosen internal volume parameter. The combined data used for these
correlations is shown in Figure 3, Figure 4 and Figure 5. In the model, values of internal or
representative volume for the various components w ere calculated during execution of the design
routines for the system performance sub-model (Section 2.3.1).
47
100
upper lower limits
for valves
10 -
ou
sr
Vi
S
so
H Opipe
□ valves
0.01
0.00001 0.0001 0.001 0.01
Represettadve volume (m3)
Figure 5: Data fo r pipe and valve mass
Table 3 also contains approximate proportions of the different material compositions for each
component. The composition breakdown is obtained from product catalogues, although
compressor data was obtained from physical measurements by Garland (2004). Other
construction materials such as plastics and elastomers have been omitted because the proportions
are relatively small compared to metals. Correspondingly, their GHG emissions are negligible,
although if emissions related to other environmental impact were to be considered, the
contribution may be significant (Frischknecht, 1999).
Table 4: Emissions resulting from raw material manufacture
CO; equivalent emissions per unit mass of material. ecpnt (kg CO: kg"1)
Material Gover et Norgate and nun. mean,

SAEFL Frischknech
Pira (2001) al Rankin max
(1998) t (1999)
(1996) (2000)
Aluminium 7.64 2.06-6.56 1 5.99 - - 2.1, 5.6, 7.6
Brass 2.97 - 11.4-16.1 ! 3.0. 10.2. 16.1
Copper - 5.54 13.2 - 3.3-6.2 T 3.3, 7.1, 13.2
Steel/iron 2.95 1.60-2.78 * 5.46, 1.79 - - 1.6, 2.9, 5.5
Oil - - 1.6 - 1.6. 1.6, 1.6
1 Low- and high-recycled content. f Different manufacturing route.
The mass data must be converted to values of equivalent carbon dioxide emissions due to the
production and manufacturing process. This was addressed by sourcing emission and energy data
for each primary material, and an additional factor is applied to account for the construction and
assembly phase. Emissions data was found in several different sources and presented in Table 4.
A mean value based on these was used in the model. Where data for recycled material was not
48
available, it was reduced by a factor of 0.4, which was obtained from the difference between
recycled and non-recvcled materials from Pira (2001). The mass o f C 0 2 emissions due to the
production of a component (M'cpnl) is calculated using equation (13).
= Z - V’. ■«<*..«+ K ■ <13>

i j
where X is the composition (as a mass fraction) of a material and s cpnt isthe emission factor for
the component material, from Table 4. K is the factor for construction emissions and a value of
K = 3 .0 was used, which was based on data for compressors from Hansen and Gustafsson
(1996).
Table 5: Emissions during construction, assembly and installation
Stage Emission source Calculation

Transport components Vehicle (fuel) Equation (20). where g = 0.00006 kgC02km
to factory' 1 (Pira. 2001)
Assembly (or sub- Brazing/metal working Equation (17). where ebra, = 0.031 kgC02 kg'
assembly) 1 of system materials (Frischknecht 1999)
Pressure/leak tests Hose/manifold and trace See refrigerant handling Section 2.3.3
system losses
Transport equipment to Vehicle (fuel) Equation (20). where g = 0.00006 kgC02km'
site 1 (Pira. 2001)
Assembly Brazing/metal working Equation (17). where ebra_ = 0.031 kgC02 kg'
Personal transport 1 of system materials (Frischknecht, 1999)
See Transportation Section 2.3 .3
Pressure/leak tests Hose/manifold and trace See Charging Section 2.3.3
system losses
Charging Hose/manifold and losses See Charging Section 2.3.3
2.2.2 Construction, assembly and installation

The construction and installation process is characterised by a combination of many different
activities, resulting in several different sources o f GHG emissions. Typically, components will be
transported to a factory' where the equipment will be constructed to some degree. Pressure and/or
leak tests will be made. The parts will then be transported to the installation site and the
remaining components will be added to the final assembly. The equipment will normally be
subject to a full leak test, and possibly repeated until the system integrity' is ensured, and the
sy stem will receive a full charge. For smaller systems (e.g., appliances), this will occur only at
the manufacturing location. These sources are therefore included in the calculation o f refrigerant
49
( M 'cnsJ) and CO: ( M ' cnsJ) emissions for construction activities, equation (14) and (15).
respectively.
(14)
M L, = I ' ' t ( 15)
The use of brazing torches produce emissions, and the amount of gas used is related to the
material mass for a given installation (equation 16). From Frischknecht (1999). values ranging
from 04)18 to 0.043 kg CO: per kg of system materials were obtained. The number of visits
requiring personal transport is estimated at one person-day per 100 kg of system mass, based on
discussions with contractors (Broadhurst, 2002) (equation 17).
(17)
where l,frav is the total distance travelled in km. for the total mass of the system components, and
g is the emission from fuel for transportation per distance travelled.
Start.
S\ Mem design
Compressor operation Ambient Vmperatures
Heat load
l .J ,.i n c e eomptM e ,v Refrigerant loss
Fouling, wear
t Visit frequency
Servicing
Technician behaviour
Figure 6: Overview o f in-use emission sources and influencing parameters
2.3 In-use stage

In-use stage emissions occur primarily from energy production for the equipment operation and
the leakage of system refrigerant charge, although refrigerant emissions associated with serv icing
activities may also be significant (equation 18 and 19). Figure 6 summarises the sources and
some key causes o f GHG emissions during in-use stage.
ML =M:„,r +AC„. (18)
M L = M ’optr+M L (19)
2.3.1 System refrigerant emissions

The annual system leakage was estimated from system refrigerant mass using equation (20).
M'oper= R oper- M rdes (20)
where Roper is the annual leak rate from the system during operation and M r des is the “design"’
refrigerant charge. The design charge is calculated based on the internal volume of the system
components, and local refrigerant density at the design, or rated conditions (as described for the
model in Section 2.3.2).
Flare Joints
K.DM Valves
Clcck valves
0 5 10 15 20 25 30 35
Figure 7: Proportion o f refrigerant leaks by mass and number o f visits (Radford. 1998)
Considerable work has been conducted to establish typical refrigerant leakage rates for various
RAC sectors. Efforts have been made to confirm these data by measuring atmospheric
concentrations to estimate the annual and total emitted mass o f a given refrigerant. Whilst this
technique has been shown to correlate well w ith total estimated emissions for a given species
(McCulloch et al, 2003), it does little to help estimate leakage from individual systems. Instead,
characterising leak rates from individual components can provide a more accurate indication of
leakage from specific configurations. Radford (1998) presents data on leakage quantities from
components in commercial refrigerating systems. Using this and more recent data for the same
and different systems from Ayers (2000), a factor ( x ) was derived by dividing the proportion of
loss by the average number o f each component in a typical system from where the original leak
data was obtained. The factor allows calculation o f annual leak rate for any set of components
(equation 21). Here, the first term accounts for the rise in annual leak rate due to the age o f the
system, and the second term calculates the initial leak rate from the combination o f components
adjusted for repair frequency', and AR accounts for leak rate uncertainties.
where d R / dt represents the incremental increase in annual leak rate for the system over the
reference time (At'). N cpnt is the number of components of a specific type, and N xn. is the
number of visits per y ear by a service person to check for and repair leaks. The component leak
index ( / ) was determined for each component type. AR is the mean variation in annual leak rate
based on the literature, which is discussed later in this section. (Note that the theoretical limit of
equation 21 is Roper * N xn, < 1 so that the sy stem has a positive charge size at any time.)
Table 6: Refrigerant leak quantity and leak factor for specific components
Total mass Total mass Component leak

Average no. of
Component released (Radford. released (Ayers.
components factor, x (>’')
1998) (kg) 2000)(kg)
Coil 7071 907 120 7.304 x io 4
Pipeing/brazed joint 35576 4563 1700 2.594 x io 4
Control valves 7918 1016 3 3.271 x io-:
Check valves 5794 743 3 2.394 x 10'2
Ball valves 1948 250 288 8.384 x 10'5
Service valves 16655 2136 60 3.441 x 10-3
EPR valve 8876 1138 5 2.200 x 10’2
Solenoid valves 3734 479 5 9.257 x 10’3
Expansion valves 6369 817 120 6.579 x lO"4
Control lines 5221 670 50 1.294 x 10'3
Control equipment 1145 147 50 2.838 x 10-4
Filter 1553 199 4 4.812 x 10‘3
Compressors 3995 512 12 4.126 x 10 3
Flare joint 3417 438 5 8.471 x 10‘3
52
Table 6 provides the systems' leaked masses and associated leak factors for a variety of
components and connection types. These data are based on values obtained from 410
supermarket systems, which received 2381 service calls throughout the year. The average number
of sendee visits was therefore 5.8, rounded to six per year. It should be noted that the correlation
provides an average value since the data from Radford and Ayers showed that the leak rate o f
individual systems actually varied from 0% to 82% per annum. The inclusion o f the frequency of
service visits in equation (21) is important because when the leaking component is repaired, a
proportion of the annual leakage is prevented. This approach is validated by the leak data from
Radford, shown in Figure 7 w here the leaked mass (top-up) is closely related to the number of
service visits to repair a given component. In terms of the incremental increase in leakage with
time, an approximate value of d R / d t = 0.054 y'1 was obtained with the data from Avers.
However, the confidence for this value is low because data was only available for systems o f up
to four years old, which showed wide scatter. Radford (1993) claims that both refrigerant and
lubricant type have been shown to produce significantly different leak rates For example, HFCs
tend to produce higher leak rates than CFCs because of their molecule size and the greater
“scouring" ability of polyolester oils. However, since the data in Table 6 is a compilation o f
various HFC and HCFC refrigerants the correlation does not allow for refrigerant-specific leaks.6
Table 7: Comparison o f reported annual leak rates
Reported Roper (Source) Equation (21) ( r = 0)

Sector/
equipment Johnson March Haydock ETSU Clodic D
(1998) (1999) (2003) (1997) (1998) ^ cpnt oper
Domestic 1% 1% .3-0.7% 2.5% .7-0.9%

8-15 1/15
ref.
Retail ref. 9-23%
- Integrals 1% 3-5% 2.5% 10-30 1/3 .8-4.3%
- Split/dist 10-20% 8-15% 15% 20-50 1 4.-12.4%
- S/market 10-25% 10-20% 8% 16-31% 1-3000 6 13.5-28%
Air cond. 12-20%

- Split/dist 10-20% 8-12% 20-80 1 4-17.7%
- Chillers 15-22% 3-10% 3-5% 4% 2-20% 20-40 2 3.1-6.0
6Although the data includes losses from refrigerant handling, the relative mass is very small compared
to system charge and is therefore neglected for the leak factor.
53
Verification of this new approach was achieved by comparing equation (21) against data obtained
by measurement or estimated leak rates for specific types of equipment. Table 7 summarises leak
rates found in the literature, and the results o f equation (21). It was found that calculated values
of R oper were generally within the ranges of those cited in Table 7. Referring back to equation
(21), the values for AR are obtained from the ranges of leak rates reported in Table 7. Across all
equipment types, minimum AR = 0.43, mean AR = 1 and maximum AR = 1.91.
2.3.2 System energy use emissions

A significant proportion of the GHG emissions can be related to the energy' consumption of the
equipment throughout its in-use stage. The energy -related C 0 2 emissions are calculated from the
equipment energy' consumption (equation 22).
<22>
i
where £ is the “emission factor”, E is the energy' consumed by the system, and i represents
each set of conditions.
upper low er limits for j

emission factor !
0.8
C
U 0.6
so
I 0.4
O historical
a
“ 0.2
C
o
0.0
1965 1985 2005 2025
Year
Figure 8: Historical and projected power station CO2 emission factor (DTI. 1998, 2000)
Within the UK, electricity' is generated through a number of different routes including coal, gas,
oil, nuclear and a number of renewable sources, each of which results in different levels of GHG
emissions per unit o f energy produced.7 The average annual kg C 0 2 per unit of electricity' (termed
“emission factor”, £ ) changes with the introduction of new , or withdrawal of old pow er stations.
Similarly, the “fuel mix” changes according to total demand at that time, and therefore the
7 Note that in countries such as Norway where nearly all electricity' originates from hydropower, the
power station emissions component is negligible.
54
emissions factor varies. Such changes in emissions factor should be taken into account because
these changes occur throughout the equipment life, and different types of system may operate for
different periods throughout the day and year.
Table 8: Constants fo r emissions factor projections
Projection scenario
Constant
Lower Mean Upper
a\ 475.12 502.59 583.01

a2 -0.46972 -0.49632 -0.57566
a3 0.00011620 0.00012263 0.00014219
VY 0.0567 0.0631 0.0693
Figure 8 shows how the average annual emissions factor has reduced over the past 25 years
(DTI, 1998). Also included are the projected emissions until 2020 (DTI, 2000), where the two
data sets represent two alternate economic scenarios. Data for the two scenarios was fit to an
equation for use in the model (equation 23). In addition, seasonal and dav/night variations are
also accounted for, following the approach of Ure (1995) and Beggs (1996) respectively. As with
the production of construction materials, electricity generation produces emissions of GHGs other
than C 0 2, but have been converted to C 0 2 equivalent using GWP(100) for purposes of model
simplicity.
£ = (a, + a2 Y +<73 •T2)+(vj. •sindY ) + (23)
where the first term represents the mean annual emission factor (see curve fit in Figure 8), ,
a 2 and a 3 are constants and Y is the year, e.g., 2005. The second and third terms represent the
annual and daily variation in emissions respectively, where vY is the mean deviation of C 0 2
emissions throughout a year, and vD is the mean deviation for day ( v D = 0.088 kgC 02 kW '1h'1)
or night ( vD —- 0.088 kgC 02 kW '1h*1). 0Y is calculated according to the peak emissions over
the annual time frame, using equation (24). Values for a x, a 2, a, and vY are presented in Table
8, and were obtained from DTI (1998) and DTI (2000).
-9 .1 0 7 ) (24)
Y ,Tot
55
where tr Tot is the total number o f time increments in the cycle (i.e., 12 months), tY is the time
(in months) and the constant 9.107 is the time at which peak % occurs.
Refrigerating system performance

Due to the potentially high emissions from energy production, it is important to assess system
energy consumption in detail. Since comparisons are often made between refrigerants, which
perform differently according to the conditions, the model should have a strong dependence on
refrigerant properties and should account for transients such as changing ambient temperature
and effects of ageing. For any set o f conditions, the performance of the refrigeration cycle is
calculated and the associated energy’ consumption is used to estimate the C 0 2 emissions from
electricity production. Also, since calculation o f performance demands estimation o f the systems’
charge size, this was also used for determining refrigerant emissions elsewhere (Sections 2.2.2,
2.3.1, 2.3.3 and 2.4).
It is noted that the performance model was not validated, for two reasons. First, no test data was
available where measurements accounted for the various parameters included, such as ageing
aspects. Second, the purpose of the model is for comparative purposes and not for prediction of a
particular piece of equipment. Nevertheless, the majority of the sub-models are largely empirical
which provides confidence in a realistic output.
The model was developed for a single-stage vapour-compression cycle, and was used for five
types of performance calculations:
(i) Steady system performance at design conditions. Based on specified design conditions, the
compressor size (swept volume) is determined, according to the condenser, evaporator and
pipework design, condenser heat transfer fluid (HTF) inlet temperature, application
temperature, the temperature surrounding the application, and refrigerant selection.
(ii) Steady system performance at non-design conditions. Based on specified component
designs and the compressor size determined for (i), performance is calculated at non-design
ambient conditions, whilst assuming that all dependent temperatures are allowed to float.
Non-design conditions such as different condenser HTF inlet temperature and/or
56
temperature surrounding the application, due to monthly mean day and night temperatures,
used to calculate energy consumption over a 12-hour period for each.8
(iii) Non-steady system performance at non-design conditions. Based on specified component
designs and the compressor size determined for (i), performance and 12-hourly energy
consumption is calculated for the non-design monthly mean day and night ambient
conditions. However, application temperatures are not allowed to float outside a specified
range so the system operation is controlled to ensure the application temperature is within a
chosen range about a set-point, and this may produce on-off cycling of the compressor. If
the ambient conditions are such that the system cannot provide the capacity to achieve the
low er application temperature, steady operation occurs as (ii).
(iv) System performance at design and non-design conditions with ageing. With each time
step throughout the equipments in-use stage, ageing occurs in the form of degradation of
component performance and loss of refrigerant. Over time, system capacity deteriorates
which affects system equilibrium conditions for steady operation (ii), or cycling behaviour
for non-steady operation (iii), and therefore energy consumption.
(v) System performance with retrofitted refrigerant. Based on specified component designs
and the compressor size determined for (i), performance is calculated according to the
properties of a different refrigerant chosen as a retrofit. Thereby the performance of the
system under (ii) steady, (iii) non-steadv, or (iv) ageing situation can be determined.
Calculate steady-state performance at design conditions, determine compressor size (i)
Existing refrigerant Retrofit refrigerant (v)
Without ageing With ageing (iv) Without ageing With ageing (iv)
JZ H I __
Non Non- Non- Non
Steady- Steady- Steady- Steady-
steady steadv i steady steady
state (ii) state (ii) state (ii) state (ii)|
(iii)' (iii) (iii)' (iii)
Figure 9: Options fo r different types o f performance calculation
8 A more accurate “bin” type approach was considered, but this would conflict with the inclusion of
ageing aspects; although, monthly “bin” data could combine with ageing models, but would result in
excessive calculation time.
57
The various options for performance calculations are indicated in Figure 9. Initially, the design
calculations (i) are necessary so that compressor size is obtained. Once the purpose o f the
simulation has been chosen, the appropriate sub-models and calculation sequence are selected,
which include existing or retrofitted refrigerant (v), ageing (iv) or no ageing, and non-design
conditions under steady operation (ii) or with on/off cycling (iii). For any of these off-design
calculations, a general iterative sequence w as employed to determine equilibrium states for steady
operation, or the performance at non-equilibrium conditions betw een the prescribed upper and
lower application temperature limits. The general calculation sequence covering all situations is
described in tw o steps: first for the determination of compressor size (i), and secondly for
performance at non-design conditions (ii, iii, iv and v). Details for the various component sub
models are provided below, as are the methods for evaluating transient influences such as
compressor w ear, heat exchanger fouling, and effect of leakage.
For the first step o f calculations, a number of input parameters are required which include design
temperatures, characteristics o f the application heat load and equipment construction details,
whilst the compressor size is determined by the model so that its capacity7is matched to the heat
load at the chosen design conditions. Initially, the saturated compressor suction (Ts ^ ) and
discharge ( Ts disch) temperatures are estimated from the application temperature ( Tapp), which in
this case is the design ( Tappdes) and ambient {Tamb) temperatures (based on a 10 K approach).
For a given space to be cooled, the application heat load ( Oapp) is obtained from a basic rating
equation (25).
QaPP{Tapp) = b, + Bapp • (Jamb - T app) (25)
where bx is a constant and represents heat loads that are independent of ambient temperature,
such as electrical devices and occupants. Bapp is the application basic rating and Tamb is the
temperature surrounding the space. In the case of a refrigerated box Tamh is room temperature, or
for an air conditioned space Tamb is the outside temperature, and Tapp is the application
temperature, inside the cabinet or the room respectively.
From this the mass flow o f refrigerant ( mr ) is calculated from the specific enthalpy difference
based on Ts suct and Ts dlsch (equation 26).
58
m= Q° " (26)
H , - H a
Next, mr is used to determine the capacity o f the evaporator (O e ) and capacity of the condenser
(O c ) from the evaporator and condenser sub-models discussed below, as well as refrigerant
outlet states for calculation o f the succeeding component. Heat transfer and pressure losses
associated with interconnecting pipework are also calculated, which are needed for the
compressor suction and discharge and TEV inlet states. Using the evaporator and condenser
characteristics, revised values for refrigerant states at the compressor suction ( ) and
discharge ( Tdisch, p dlsch) points are determined. mr is then re-calculated from equation (26),
where H x corresponds to , p ^ . O e and O c are subsequently calculated again, and the
process iterated until Oe = Q , where the system capacity matches application heat load for the
design conditions. Henceforth, the design compressor size (in terms of swept volume, ) is
found (equation 27).
z. i d t / y* t1 \
y _ 1 &PP *' omb app.des * (Jl\
• P suct ' ( H \ ~ H 4)
where the upper term represents the application heat load, and therefore evaporating capacity'
(equation 25), tjw! is the volumetric efficiency and is refrigerant density at the compressor
suction. V^, was converted to cy linder volume (assuming 1400 rpm compressor speed) for
calculation of compressor and oil mass in Section 2 .2 .1.
For the second step of calculations, each set of non-design conditions (ambient temperatures,
ageing effects, refrigerant, etc.) are specified whilst maintaining existing system design with the
previously determined compressor size (equation 27). Since Oe is now' unlikely to match Oa at
the required Tapp des, it is necessary' to establish sy stem performance characteristics at the new
sy stem equilibrium conditions, or at incremental T between the upper and low er limits
specified for the application temperature. In this case, incremental Tapp is interpreted as a floating
application temperature, denoted T'app. Determination o f performance characteristics for any of
the calculation types (ii, iii, iv, v) follow s the same procedure.
59
Firstly, an initial guess of TSSUCJ = Tapp des - 10 K and Ts dlsch = Tamb + 10 K is made as before,
and mr is now estimated from (equation 28).
(28)
With mr , the performance o f each component (Q e , Oc , heat transfer and pressure losses
associated with interconnecting pipework) is calculated. Oe is then compared against Oapp, and
if they do not match, T'app is adjusted incrementally ( S T ') until Oe matches Oapp at the new
application temperature ( T'app ± S T ' ). S T ’ for each new iteration is estimated using values
from the prior iteration in equation (29), which is based on the relationship between change in
application capacity with temperature. It was found that this convergence approach (as opposed
to say, successive guesses of mr proportional to Oapp I Qe ) was necessary’ to avoid instability.
' aPP v aVV>

nPP- s r
6 T ' = exp (29)
O
i - e (T ' )“
' app j KQapp(T'app) - S O 'appJ
where Oapp {T'app) and Oe{T a' pp) are capacities at T'app from the previous iteration, and S Q'app
and S T ' are the incremental change in application heat load and temperature respectively, also
from the previous iteration.
If the model is set for steady operation (ii), then T'app is adjusted until Oe{T'app) = Oapp {T'app),
and energv’ consumption is calculated on the basis that the system is in constant operation. If non-
steadv or cycling operation occurs the calculation of energv consumption is determined according
to the method described below.
System cycling
It is important to know’ compressor cycling time, as this has a direct influence on power
consumption (although power surge at compressor start-up has been neglected), and indirectly, on
degradation of compressor performance. In order to evaluate cycling time, a simple method w as
developed as follows, based on application heat load characteristics and system performance at
off-design conditions, but neglecting pull-down from the initial (ambient) temperature. Cycling
occurs due to termination and initiation o f a single cylinder compressor, in response to a
thermostat which is set to respond to an upper ( Tappmax) and lower ( Tappmin) application
temperature, whilst the actual temperature ( T 'app) floats between the two; towards Tappmin during
60
compressor on-cvcle and towards T max during off-cycle. The difference between Tappmax and
Toppmin is the thermostat setting ( ATset), and the design temperature is the mean, Tapp des -
Tapprmn + ATset / 2 . It is noted that any lag due to the thermal response of the thermostat itself is
neglected in this assessment.
As mentioned above, is calculated for design conditions so that Oe meet the design load of
the application (at full charge and no ageing), thus at design conditions the compressor operates
continuously and T' = Tapp des. It is assumed that the design conditions represent those that
would produce (at least) the greatest demand on the refrigerating system, so under off-design
conditions, the system has ‘‘excess” capacity. As a result, once T des is achieved Oe exceeds
Qapp{Tapp des), and T'app continues to fall from Tapp des towards Tappmn. In order to calculate the
compressor (and ancillary devices) energy consumption, it is necessary to determine the average
“instantaneous” power demand at each incremental T' , and corresponding pull-down time
(between Tappmax and Tappmm) for other on/off electrical devices.
The rate of decrease in application temperature - or pull-down rate - during the on-cycle is a
function of the thermal mass of the contents of the cooled space, and the excess capacity o f the
refrigerating system that enables a lower temperature to be achieved (equation 30).
^
dt (M • Cp) app
where A0(T'app) is the difference between Oapp(T'app) and 0 €(T^pp) , and (M Cp)app represents
the thermal capacity' of the space.
For each step o f T a' pp, pull-down time for each ST' can be obtained by rearranging equation
(30), and substituting AQ(T'app) for Oe(Tjpp) and Qapp{T'app) from equation (25), as equation
( 31).
(M ■Cp)app
dt = -----------------------------
/rryf \ i r% /m nnt \
d Tapp
' (31)/
V
Qe(T'app) - b x + B app-{Tamb-T 'app)
Total pull-dowm time - or on-cycle time ( Atonc) - requires equation (31) to be integrated over the
T'app range according to the thermostat setting ATxel (equation 32).
61
start
4- (M • Cp) app / AO(TqPP ) •dT'app (32)
where Atstart is an additional term which accounts for the extra time at start-up due to degassing
of refrigerant absorbed in the oil and flooding of refrigerant into other parts of the system. Whilst
this value is dependent upon oil quantity, refrigerant/oil properties and external conditions, a
single time of 90 s has been used, as determined by Phillip et al (2000). However, this is known
to vary considerably in different types of systems.
Similarly, the off-cycle time ( Atofc) is calculated as with equation (32), but since the refrigerating
capacity- is absent, the change in T',PP is dependent only upon the application heat load, and the
Atstar1 term is not relevant (equation 33).
(33)
As required for certain ageing calculations, the number o f compressor starts ( N start) over a given
period of time ( A t ') is determined from equation (34).
N = ------—------- (34'
"afT ,
A/ one + At oje
For presentation of results, Tapp, Oapp, Oe , COP, etc., are given as time-averaged mean (based
on &onc and M ofc) as T'app floats between Tappmax

app, max and T.
System component sub-models

Sub-models for the main system components - compressor, evaporator and condenser - were
those described elsewhere in this study (Part 2, Section 3). However, some minor modifications
were introduced to address certain factors relevant to the current purpose.
Compressor performance is highly variable due to the large number o f designs and combinations
of refrigerants and oils, and to account for such a wide number of parameters would demand very
detailed methods and thereby deviate from the current purpose. Whilst the intention is partly to
assess differences between refrigerants, it was argued by compressor manufacturers that it is
possible to optimise a compressor to suit any given refrigerant and oil combination (Jurgenssen,
2004, Blomhardt, 2002). Therefore, an approach was chosen where such effects are eliminated
62
from the direct calculation of compressor power consumption (although the existing method for
calculation of refrigerating capacity and volumetric efficiency are still employed). Subsequently,
the compressor model presented elsewhere (Part 2, Section 3.2) was adapted so that instead of
calculating compressor power ( Wcomp) with the ideal mean pressure approach, it was determined
from compressor refrigerating capacity ( Qe comp) and an empirically derived COP (equation 35).
w = Qe,comP (35.
comp C O P
comp
where COPm is the coefficient of performance calculated from an empirical correlation based
on the performance o f a real compressor as a function of suction ( ) and discharge ( p dlsch)
pressures only (equation 36).
C O P oomP = exp (tf, + a 2 ■p MCt + a 2 • p ^ + a4 • p wef • p dtsch) (36)
where a x, a 2 and al are obtained for a specific compressor, and exemplary values from
catalogue data for tw o different products are provided in Table 9, and are an average of the
various refrigerants for which data was provided. By calculating Wcomp in this manner isentropic
efficiency is dependent mainly upon the compression ratio, and therefore system COP is
ultimately dependent upon the performance of the respective refrigerants within the other
components, such as heat exchangers and pipework.
Table 9: Constants fo r equation (36) fo r two different compressors
Compressor (manufacturer) 2 KC-05.2 (Bitzer) 2 GC-2.2 (Bitzer)
a\ 8.374 x 101 8.048 x 101
a2 3.239 x 10*3 3.469E x 10 3
ai -5.826 x KT4 -5.021 x 10'4
a4 -5.334 x IQ'7 -7.625 x 10'"
The total energy drawn by the compressor is the product of Wcomp at the given conditions and
operating time (equation 37).
E comp = /Y *W comp a -A t o n c j (37)

W 1/
i
where Atonc is the duration of the compressor on-cycle for a given set of conditions, i .
63
Finally, estimation of the heat that the compressor adds to the refrigerant is assumed to be equal
to the compressor power, minus 5% o f the heat (Lundqvist, 2004) which is rejected to the
environment (equation 38),
Qconp = ( 1 - 0 . 0 5 ) - ^ (38)
Heat exchanger performance was calculated with the evaporator and condenser models detailed
elsewhere (Part 2, Section 3.3), and the internal volume of the evaporator and condenser is used
to determine component mass for Section 2.2. In addition, the power required for HTF flow
across the evaporator and condenser was evaluated. This was considered important since a higher
HTF flow rate increases heat transfer coefficient and thereby improves the COP, whilst being
offset by the additional power required by the fen or pump. This was calculated using equation
(39), which combines the conventional proportionality’ equation and a correlation for pressure
drop across an exchanger.
1.8
f yv HTF ^
W =c•
rr U'TLT — (39)
H cpnt V A HTF J
where r\cpnt is the efficiency of the pump or fen, and c is an empirical constant (in kJ m -3\') for a
specific pump or fan in a specific HTF circuit or duct arrangement. A HTF is the face area of the
exchanger in the direction of HTF flow.
Pressure losses along suction anddischarge lines are included in the model, whilst liquidlines are
neglected since the pressure drops are known to beinsignificant. Conventional single-phase
pressure drop correlations are used (equation 40) (Incropera and DeWitt, 1990).
t y = X P r ' Ur L (40)
2 •D
where p r is refrigerant density7, ur is mean velocity, D is inside tube diameter and L is pipe
length. The friction factor is from equation (41) (Incropera and DeWitt, 1990).
A = a -R e n (41)
where Re = p r - ur - D ! p r , and p r is refrigerant viscosity7. Properties were calculated for local
temperatures and pressures at each increment. For Re < 20000, a =0.316 and n = -0.25, and
a =0.184 and n = -0.20 when Re > 20000.
64
Similarly, calculations for pipework included finding internal volumes which was used to
determine the mass for pipe and valve materials, in Section 2.2.
Electrical devices associated with the refrigerating system should also be added to the calculation.
This ancillary equipment includes evaporator and condenser fans, lighting, control circuits,
defrost, oil and trace heaters. The energy consumption o f these is simply the sum o f their
electrical rating ( WanciJ) and their respective run times ( Atonc) for each device, i (equation 42).
(42)
Transient influences on performance

In addition to the previously described sub-models, three time dependent influences alter
performance over time. With continued use of the system, degradation of system performance
occurs in the form o f compressor wear, exchanger fouling and sub-optimal refrigerant charge.
Therefore, for a refrigerating system operating under constant conditions, the COP and capacity-
wili reduce with time.
Compressor wear
Compressor wear causes a reduction in volumetric flow of refrigerant and therefore cooling
capacity. A lower cooling capacity and (at least) the same motor power results in lower
compressor COP (Garland, 2004). Whilst these mechanisms are related to operating time, most
are a strong function of the number of compressor starts. Contact surfaces are initially absent of
oil, and when oil becomes available, it is initially well mixed with refrigerant making it less
effective as a lubricant; subsequently, more surface buffing and changes in geometry occurs. In
the performance model, degradation in compressor performance is accounted for by using a
simple empirical equation for volumetric efficiency (43) and compressor COP (44), obtained
based on the data fitting of Garland's results.
" y j C p ' N s ta r t ) (43)
dCOP
‘ -\/ C p ‘ ^ s t a r t ) (44)
65
where a is the same for both equation (43) and (44); for maximum, mean and minimum
uncertainty limits, the value of a is -674.5 x 10'3, -27.0 x 10‘3, and -13.5 x 10'3, respectively.
t op is the operating time of the compressor (in years) and N ^ is the number of compressor
starts. Although Garland (2004) provides experimentally derived equations for estimating the
reduction in flow rate, it requires detailed knowledge of the internal geometry of the compressor.
This approach was considered too thorough for the current model, so instead typical values
provided by Garland (2004) and Jurgensen (2004) were used to determine a in equation (43) and
(44), for a 10% reduction in Oe for 20,000 operating hours and 100,000 starts.
10000 T
1000 -
^ 100 -
* McAKsteretal
'h
Greissetal
Gitzeit
Ritter and Suitor
0 4 8
Time (years)
Figure 10: Changes in heat transfer o f fouling over time
Heat exchanger fouling

Modifications were made to the existing heat exchanger models to account for the influence of
fouling. For condensers and evaporators, refrigerant-side fouling is normally minimal (excepting
oil effects) since the internal circuit has to be free o f contaminants, and was neglected. However,
significant fouling does occur when water, air or other secondary HTFs are used because the
conditions are relatively uncontrolled. A number of studies have addressed the issue, and most
present formulations to estimate the additional thermal resistance, i.e., increase in x l k ,
associated with the fouling. Examples of studies of fouling effects on heat transfer coefficient
over time are presented in Figure 10.
Whilst some models are very comprehensive, a more simplistic approach was used, as detailed by
Butterworth (2002), (equation 45).
66
\
d(x >k )/M —
_ a_ l Re lfTF • exp - a 2 -s„ (45)
at 8.314* (273 + J H)
where a x = 8.39 s m2 K W '1, 8.314 is the gas constant, TM is wall surface temperature and E =
68 000 J mol'1and is the chemical reaction activation energy. Whilst a 2 is provided by
Butterworth for oil, the applicability o f equation (45) was extended to heat exchangers used in
refrigeration by correlating to empirical data reported by Somerscales (1997), producing the
values in Table 10. s M, is the wall shear stress (equation 46).
S w = C • R C tfjp • P ht F 'u (4 6 )
where c and n depend on the type o f flow (Incropera and DeWitt, 1990). For flow over fins, c
= 0.332 and n = -0.5 for laminar, and c = 0.664 and n = -0.5 for turbulent flow. For flow
within tubes, c = 0.158 and n = -0.25 for laminar, and c = 0.092 and n - -0.2 for turbulent
flow.
Table 10: Constants fo r fouling equation (45)
Source Materials a 2 (Pa s n r K W 1)

Butterworth. 2002 Steel/oil -4.03 x 10'11
McAllister et al (Somerscales. 1997) Copper/water -5.44 x i o 13
Greiss et al (Somerscales, 1997) Aluminium/water -1.85 x i o 12
Gutzeit (Somerscales. 1997) Aluminium/water -2 . 0 6 x i o 12
Ritter and Suitor (Somerscales. 1997) Copper/water -2.69 x IO’12
Lower - -5.44 x i o 13
Mean - -1.79 x i o 12
Upper - -2.69 x i o 12
The additional thermal resistance due to fouling after a given duration of operation is accounted
by inclusion of the calculation o f overall heat transfer coefficient (e.g., equation 47). It was
assumed the fouling was equally distributed along the entire HTF surface.
1 1
+ + ---------
3r*
hr
i
where hr and hHTF are the refrigerant and HTF heat transfer coefficients respectively.
67
Leakage effect on performance
When refrigerant leaks the total quantity o f refrigerant within the system reduces and thereby
starves the evaporator. The resulting effect is increase in temperature lift and loss o f cooling
capacity, which forces an increase in run time and generally a reduction in COP. The
consequence o f leakage was integrated into the model by simulating the operating characteristics
of an under-charged system. Since lowered refrigerant charge manifest as cessation o f phase-
change progressively further away from the evaporator outlet, it was simulated by reducing the
surface area available for phase-change ( A tp) m some proportion to the loss o f charge (equation
48); the remaining evaporator area provides superheating only.
A ^ (<p)
= «*(f)< 1 (48)
Av
where (p{t) is the reduced charge due to leakage at time t and n represents the sensitivity7o f the
evaporator to charge reduction. The reduced charge can be expressed as the charge mass at t as
a fraction o f the design charge (<p(t) = M r ( t ) / M r des), or as the “unleak” fraction {(p{t) = 1 -
Roper ( f) ), and M r des is the optimum refrigerant charge for the system under design conditions.
Table 11: Studies on effects o f refrigerant leakage
Aprea and Galloway and

Farzad and
Source Mastrullo Goldschmidt Model
O'Neal (1992)
(2002) (1993)
type Non-specific Split a/c Split a/c
Refrigerant R407C Not known R22 R22
Evap/cond Air/air Air/air Air/air Air/air
Exp. device TEV, EEV TEV, Cap TEV, Cap TEV
Capacity' 3.5 kW 12 kW 11 kW 11 kW
Charge size 4 kg 3.6 kg 4 kg 3.0 kg
Ambient 18°C Not known 35°C 35°C
Room temp 26°C Not known 21°C 21°C
To determine realistic values of n , the system model was compared with data from five different
studies, details for which are listed in Table 11. Information about any o f the tested systems was
insufficient to make a direct comparison against the model because many system characteristics
were absent. Instead, simulations wrere made using inputs typical of an air-to-air split air
conditioner/heat pump (representative of most of the equipment in the studies). Some results
based on the conditions in Table 11 are plotted in Figure 11. By varying values of n , it was
found that nearly all of the reported variations in COP fell within the range calculated by the
model w hen n = 1/3 and 3/2. These were chosen for the upper and lower limits, with n = 3/4
being the mean. Two cases lay outside this range; the data from Farzad and O'Neal (1992) for a
TEV system implies that the optimum charge was lower than specified, hence an increase in COP
with loss of charge. If the data were compared relative to an optimum charge 10% lower, then the
data would lie within the modelled range.
Since performance must be evaluated progressively throughout the equipment life, it was
necessary to determine the charge deficit at each time increment. It was assumed that the leak rate
was constant, so that the reduced charge (p{t) is linearly proportional with time, and that service
calls are spaced equally over a year (with the final visit at the end o f the year). At each service
visit, the service person finds the leak and replenishes the sy stem to its design charge, and since
(p{t) = 1 - Roper (t) , the reduced charge at any time throughout the year is estimated from
equation (49).
(49)
where Roper is the total leak fraction over one year, N tpup is the number of service calls per year
where the system is topped-up with refrigerant, and A t'is the reference time of one year.
1 .1 1
modd, n = 3/
/
\
Aprea et al TEV
V °-7i - o - Galloway et al TEV
Galloway et aL cap
model, n = 1/3 -o —Farzad et al TEV
♦ - Faizad et al cap
0.5 0.6 0.7 0.8 0.9 1.0

M r t < P < l ) / M rdes (-)
Figure 11: Comparison o f leak effects on COP
69
2.3.3 Emissions from servicing and maintenance activities
During servicing and maintenance, emissions predominantly originate from refrigerant handling
( M 'en.. equation 50). whilst CO: emissions ( M 'en.) are minor contributions from transport and
possibly assembling replacement components (equation 51).
M 's e n - = M ' ckrg + M "vem + M ptsi + M mem* + M 'evac (3 0)
M l =M '+M ' + M' (51)
Start
Is leakage the problem Repair leak?
I* .‘f t je ra n t vented?
Possible to pump-down
i T im p - u > « i . i . 'e 'T l
j ■Isolate com ponent for replacem ent

Y repair
Replace-'!. p. ir .t n v'jnent
N itrogen flush «.vut,u<itc svsteni
Is refrigerant leaking?
End
Figure 12: Activities producing emissions during servicing
70
Refrigerant handling and other peripheral sources
Throughout the in-use stage of the RAC equipment, regular servicing (and maintenance) is
normally provided, and has several purposes: confirming system operation, leak checking,
topping-up refrigerant, leak repair, and replacing failed components. During several of these
activities, refrigerant emissions may occur because o f transferring refrigerant. It is important to
account for these losses because studies show they can be significant. For example, Clodic (1998)
reports that 60% of emissions from chillers and 16-29% of emissions of supermarket systems
occur during refrigerant handling. Similarly , Butler (1993) reports that 30% of emissions are
from refrigerant handling activities. In addition to releases o f refrigerant, other peripheral GHG
emissions occur, such as from transportation to and from the installation. Typical servicing/
maintenance procedures are identified in Figure 12, and the specific activities that result in
emissions are identified with a shaded box.
The actual emissions associated with servicing activities are strongly dependent upon the
behaviour o f the service person. Table 12 show s the basis for the calculation of refrigerant mass
emitted during a poor and best practice servicing activity, at the stages identified in Figure 12.
Each of the sources was modelled as detailed in the following sections.
Table 12: Details o f servicing activities leading to refrigerant emissions
Basis of mass of refrigerant emission

Activity Process
Poor practice Best practice
Topping-up Charging Liquid in hoses/manifold at Vapour in hoses/manifold at
system pressure system pressure
Repair leak or Venting Release all refrigerant, Release no refrigerant Pvent = 0
component P
1 vent = 1
1
Recover Vapour in hoses/manifold at Release no refrigerant

refrigerant sy stem pressure
Evacuate Vapour in system volume at 1 Vapour in system volume at 30
system atm. kPa
Purge system 2 x system volume at 1 atm. 1 x system volume at 1 atm.
Charging Liquid in hoses/manifold at Vapour in hoses/manifold at
system pressure sy stem pressure
Venting
For various reasons, refrigerant may be vented from a system instead of being recovered. In the
model, it is assumed that venting results in the entire refrigerant charge being emitted, and no
residual refrigerant in the system, i.e., for a single system M'vem = M r des (equation 52).
71
(52)
where P^nt is the probability that the refrigerant is vented per visit, rather than recovered (see
Table 12). P^nt is interpreted as the percentage o f charge that is not recovered over a given
population of equipment, and data for typical applications is given in Table 13. Based on this
data, the average probability of venting is approximately 0.2.
Recovery
Following removal of refrigerant using a recovery machine, interconnecting hoses will still
contain vapour that is released once disconnected. Estimating the release employs the same
calculation as for charging (equation 55), but the density is of vapour at the recovery pressure,
and the probability' of not venting (1 - P^ nt) is included (equation 53).
A C *, + A C * -A C * c (53)
Evacuation
The mass of refrigerant released at evacuation stage is the sum o f the mass of vapour within the
system parts and the mass of refrigerant absorbed in oil (equation 54).
A C « = P r - ^ + A * * / - S rt/ (54)
where both refrigerant vapour density' ( p r ) and refrigerant solubility- in oil ( S oil) are evaluated at
room temperature and the pressure at which recover}' was terminated (see Table 12). The mass of
oil is approximated from equation (12) in Section 2.2.1.
Table 13: Summary o f reported venting probabilities
Source
Sector/ equipment
Johnson, 1998 March, 1999 Haydock, 2003 Clodic. 1998
Domestic refrig. 50-75% 20-50% 3-65%
Retail refrig. 21-43%
- Integrals 20-50% 10-50% 3-6%
- Split/distrib 5-10% 5% 3-6%
- Supermarket 5-10% 4-5% 3-6%
Air conditioning 15-22%
- Split/distrib 5-10% 4-8%
- Chillers 3-5% 4-5% 0-100%
72
Refrigerant charging
Refrigerant losses can occur during charging from a cylinder or a production line. The mass of
refrigerant released due to disconnection of hoses following charge is equal to that held within the
apparatus (equation 55).
M chrg ~ P r '^ h o s e (55)
where Vhose is the internal volume of the refrigerant hoses and manifold, typically 0.0007 m \
Refrigerant density p r , assumed to be at room temperature, might be in liquid or vapour phase
depending upon charging method (see Table 12).
Energy use
During service and maintenance, CO 2 emissions originate from transportation, and for the
construction and fitting of replacement components, which are calculated as SOL. Although
energy is consumed by items such as recovery machines, these sources have been ignored
presently. In travelling to and from an installation, the mode of transport will produce emissions
because o f fuel combustion. Therefore, more service calls to a site means the total distance
travelled is further and therefore higher emissions, as equation (56).
^ trav $ * ^ tr a v " ^ sen- (56)
where Lfrav is the distance travelled per visit (in km) and N sen, is the number of service visits
(i.e., journeys travelled), g is the C 0 2 emission rate for the vehicle used, which is around 0.3
kgCO^km"1 for typical petrol cars (AutoLPG, 1998).
2.4 End of life stage

At the end of its life, the equipment is dismantled, first by the refrigerant being recovered or
vented, and then the machinery’ is disassembled. Subsequently, refrigerant is sent for recycling or
incineration, as is the oil, whilst the machinery- parts (metals) are sent for disposal or recycling.
Refrigerant and C 0 2 emissions from these activities are as equation (57) and (58) respectively.
(57)
(58)
A generalised process for end of life stage is shown in Figure 13.
73
Refrigerant vented"
mmect hose&inanofold/recover
machine/cylinders
Recover refrigerant
Cancel emissions from

Refricerant recycled'’
refrigerant production
n destroyed"
-— . , ... . I !Oancel emissions from materials

M aterials reeve - *■ i ,
i production
sed"
T im cl from installation
Figure 15, Aeiivtiicr affectin g emissions during end o f life process
2.4.1 Dismantling
The mass of refrigerant vented at end of life is determined in the same way as m Section 2.33
(equation 59). Where recovered refrigerant is sent, for recycling, the emissions associated with the
production of the same mass of refrigerant are deducted from the start of life stage (Section
2.2.2). In addition, the emitted amount calculated for evacuation (Section 2 3 3 ) was assumed
equal to the amount released from the system (and oil) during dismantling.
(59)
The use of brazing and similar activities used to dismantle the equipment was assumed equal to
those calculated from equation (16) for the construction at start of life stage, and again the
contribution for transportation is included (equation 60).
M L, +AC, . (60)
74
2.4.2 Disposal of refrigerant and equipment
Releases occurring as a result o f disposal are mainly attributed to incineration of materials, such
as oil and refrigerant, which includes both refrigerant and C 0 2 emissions (equation 61 and 62).
M'Jap= Y dM ;ncmJ (61)

i
<6 2 >
i
Once refrigerant has been reclaimed, it is tested to determine whether it is of an adequate quality
for recycling or whether it should be sent for destruction, which is typically done in a toxic waste
incinerator. Normally a small proportion o f refrigerant is released un-incinerated and a small
proportion of GHG combustion product is released (UNEP, 2002). The energy input is very low
due to exothermic burning for most refrigerants. The mass of refrigerant that escapes incineration
is simply based on the plant efficiency (equation 63).
M 'men = 0“ ffincin ) "M r,incin (6 3 )
where M r incin is the mass of product sent for incineration, and rjincin is the incineration efficiency
of the plant. A typical value for rjincin = 0.997 (UNEP, 2002). The mass of C 0 2 emitted from
incineration is from equation (64).
V„ — • ittcw •A{ rr
A /‘incin lC02 + 10-<f (64)
M.
where M is the molecular mass of C 0 2 or the refrigerant and cf is the emission factor for a power
station. The first component accounts for the C 0 2 from combustion of the refrigerant, and the
second component represents the energy required for incineration, where 10 is the energy required
(kW h kg"1) according to James (1999). By comparison, James (1999) gives an aggregate figure
o f M " = 0.09 kg C 0 2 per kg of refrigerant.
Compressor oil is either put into landfill or burned. In the case o f burning the oil, the same
equation (64) for refrigerant incineration w7as used, where the molecular mass of oil (M oi7 ) is
used instead o f molecular mass of refrigerant. Typically, M o;/ = 200 kg kmol"1. If the oil goes to
landfill, the emissions are assumed zero. Other materials (metals) are either disposed o f in landfill
or recycled. As with the case of recycled refrigerant, the emissions associated with material
production at start of life stage, are altered to represent the lowest emissions factors in Table 4
(Section 2.2.1).
75
SECTION 3: CLIM ATE IM PACTS
3.1 Options for evaluating climate change impact

The introduction and literature review in Section 1 identified that the established method o f
evaluating the impact o f GHG emissions is with the use o f GWP; however, the literature revealed
that many alternative approaches exist for quantifying the global warming impact. Due to the
formulation of GWP, its application in terms of establishing the total effect of emissions of C 0 2
and the gases commonly used as refrigerants, introduces many uncertainties. In particular, this
includes the difficulty of interpreting the physical consequences of GHG emissions, and further,
the choice of a meaningful integration time for which the consequences of C 0 2 with an infinite
lifetime, and short-lives refrigerants, can be comparable. Due to these issues (discussed in in
Section 1.2.3), an alternative measure of quantifying the impact of emissions was sought. By
adopting an alternative to GWP, results from emissions modelling can be used to compare the
significance of the various emission sources against the conventional approach.
Emission Higher measureabilitv
Atmospheric concentration change
Radiative forcinj
Climate behaviour
Eco-system response
Societal impact
Economic impact Higher relevance
Figure 14: Chain o f climate relevant measures
An important consideration for the selection of any measure of GHG impact is relevance to the
consequence of the emission. Whilst the quantity o f emissions are relatively easily measured (or
predicted), the chain of events that follows become increasingly sensitive to a very broad range of
factors and so are less accurately estimated. In contrast, the emission o f GHG, the consequential
76
events following the release of a GHG have increasing significance on societies. This concept is
illustrated in Figure 14. which shows events such as changes in atmospheric concentration to be
easily quantifiable, but the ensuing changes to global climate and ultimately how this forces
society and economies to be affected, are increasingly speculative.
In this context, the concept of GWP lies around ‘‘radiative forcing'’ since it is a function of
atmospheric concentration and chemical properties. In order to shift the result of a TEWI exercise
towards greater relevance to society, it is appropriate to employ a measure associated with some
form of ‘‘climate behaviour” characteristic. Further benefit would be achieved were it possible to
overcome the essentially subjective approach of weighting long- and short-lived GHGs through
the choice of integration time (as used in GWP calculations). Section 1.2.3 identified a newly
developed index (AGTP), which characterises the consequence o f an emission in terms of a
comprehensible global temperature rise; further, such a measure also permits a damage index to
be adopted. On this basis, three new concepts were developed as alternatives to the TEWI/GWP
approach; one based on temperature rise, and two representing the subsequent societal impact.
3.2 Global warming potential

For purposes of completion, the calculation o f GWP is presented here. The GWP for any species
over a given time horizon is the absolute global warming potential (AGWP) of that species
relative to the AGWP for CO: (equation 65).
AGWPx{t)
GWP{t) = (65)
AGWPco2(t)
where AGWP is the absolute GWP at time t . AGWP is obtained from equation (66) for the
species, x , and equation (67) specifically for carbon dioxide (Shine et al, 2004).
AGWP (t) = F - t 1 - exp m

X
jr
( \
AGWPco2(t) = FC02 • a0 ■t + ]T a,; • r, • 1-exp
t
(67)
v ,
where F is the radiative forcing and r is the atmospheric lifetime of the species, x ; values for
common refrigerants are contained in Table 1. Values for ai and r, from equation (67) are from
Shine et al, as follows: a0 =0.1756, a, =0.1375, a , = 0.1858,a3 = 0.2423, a4 =0.2589, t} =
421.093y, t2 = 70.595 y, r 3 = 21.4216 y, r 4 = 3.4154 y.
77
The “equivalent"’ global warming impact of an emission of a non-C02 gas, relative to C 0 2 is
calculated from the GWP of the gas (equation 68).
M m= M'x GWP(t)x (68)
where GWP is calculated for a given t .
3.3 Global tem perature change potential

Shine et al derived equations for pulse and sustained emissions of C 0 2 and non-C02 GHGs. The
derivation was based on the additional amount of heat retained at the earth's surface (over time)
due only to the insulating effect of the emitted gas under consideration, and the duration that this
excess heat is absorbed into the oceans.9 Comparison with an energy balance climate change
model (EBM) showed that the equation for pulse emissions produced a high error, whilst the
equation for sustained emissions was within ±5% for most gases. Therefore, the equations for
sustained emissions were used for equations (69) and (70), when r ^ T .
1 < 1 - exp f o '

AGTPAt) = T- exp ------ (69)
C (T-’ - r ; 1) r I T)_
V LX )
an T t-a„ T \-exp\ - -
Fr- (70)
AGTPC02 (0 =
C'
1-exp | - - exP |----- -e x p |--
T _I - r r.
where C ' is the global surface heat capacity, which is interpreted as the specific heat of a 100 m
layer of ocean, 4.2 x 108 J K 1 m'2. T is a time constant for the climate system and is the product
of cr x C ', where cr is a climate sensitivity parameter that relates to the temperature when
doubling C 0 2 concentrations, a ranges from 0.4 to 1.2 K m2 W '1 (typically 0.8, but the
sustained emission model is relatively insensitive to a ) . F and r can be found in Table 1, and
the values of a and r (used to calculate the atmospheric “decay"’ time of C 0 2 by means of
absorption into carbon sinks) are the same as those from GWP listed in Section 3.2.
Since equations (69) and (70) are based on sustained releases, the actual contribution of a single
release is from equation (71).
9 Obviously, for a few kilogrammes of refrigerant this equates to a very small temperature rise.
78
AGTP{tY = AGTP(t) - Y ^ A G T P it,)' (71)
*=-1
where k is the number of prior years since the time of the emission.
Following the convention for GWP, the relative global temperature potential (GTP) of a non-C02
GHG is found from equation (72).
GTP(t)= AGTP‘ (,) (72)

AGTPco2(t)
3.4 Total lifetime tem perature impact (TLTI)

By applying the absolute GTP associated with the emission of 1 kg of GHG, to the calculated
quantity^ of gas emitted for each year that the equipment is responsible for a release, the
corresponding temperature rise can be calculated. This yields a novel concept, named Total
Lifetime Temperature Impact, offered as an alternative approach to the conventionally used
TEWI. How ever, the formulation of TLTI is somew hat different from that of TEWI in one
respect. Whereas TEWI is simply the summation of the product of emissions and GWP, TLTI is
a function of the maximum temperature rise experienced as a result of the total emissions. Thus,
TLTI is associated with a single point in time subsequent to the emissions, wherever the
temperature rise peaks, as equation (73).
TLTI - max Y , M , ( t ) A G T P ( t ) , + Y 1M ( l ) r AGTP(l) (f = 0->oo) (73)
where M is the mass of substance emitted: for the refrigerant, M t is the amount of each species
released, and for the contribution from C 0 2, M } is the total mass of produced typically due to
energy usage of the sy stem.
An example o f a TLTI calculation is shown in Figure 15. Total AGTP resulting from emissions
of a high global warming refrigerant and C 0 2 associated with the systems energy usage (in
existence for say 10 years), is plotted over time. The global warming impact of the refrigerant
and the additional contribution from C 0 2 is represented by the upper line. TLTI corresponds to
the maximum total AGTP, as indicated on the Figure. By adopting this approach, the impact
associated with a given refrigerating sy stem can be perceived in comprehensible units, and the
reliance on choice of integration time is also avoided since the time dimension is implicit within
79
the TLTI value. It is important to note that TLTI is not “equivalent'’ to TEWI, in that they do not
simply adopt different units. In particular, TLTI brings out the rapid but brief warming effect of
refrigerants that have a high radiative forcing but for a limited duration, whilst relying less on the
minor wanning effect that C 0 2 may have in hundreds of years in the future.
TLTI
GDI I
time
Figure 15: Illustration o f TLTI GD11 and GDI2 interpretation
3.5 Global damage index (GDI)

Further to the AGTP concept, characterising the rise in temperature makes it possible to address
the global w arming impact of emissions at a more relevant point further along the range of
climate impact measures (as Figure 14). Since the consequences of global w arming are closely
related to global temperature change, it is feasible to use AGTP or TLTI to evaluate societal and
economic damage. Within the literature review a number o f definitions for the concept termed
global damage index (GDI) or potential, were identified from Smith (2003), Fuglestvedt et al
(2003) and others. “Damage” typically refers to the economic burden on society in order for the
population to maintain its quality’ of life, at some point in the future. As explained in Section 3.1,
such a concept is significantly more theoretical, producing results o f less certainty , than preceding
measures.
For the first GDI, the concept of Fuglestvedt et al (2003) was adopted, and the damage index is
calculated from the total temperature rise resulting from the emissions over a chosen integration
period, referred to here as GDI 1. GDI 1 is calculated by integrating the total AGTP to some
exponent to the year that is of interest (equation 74), and Fuglestvedt et al suggested the exponent
be between n= 1 and n= 2.
80
GDI 1 - AG TP” dt (74)
J*
Jo
In this form, GDI 1 represents a reduced total temperature rise contribution of the emissions
within the chosen period, and is shown in Figure 15 (where n = 1.5) where GDI 1 is indicated by
the hatched area. On this basis, GDI 1 is comparable to GWP (and therefore TEWI).
The second GDI was established using a different approach where the degree o f damage is
quantified according to the rate at which global temperature rise occurs, as proposed by Shine et
al (2004), Fuglestvedt et al (2003) and others. By interpreting this simply as the average gradient
associated with temperature rise over time, an alternative GDI 2 may be quantified according to
the time taken to reach TLTI (equation 75).
TLTI
G D I! = ----- — ------ (75)
t(TLTI) - 7(0)
where TLTI is from equation (73), t{TLTl) is the time (year) that TLTI (or maximum AGTP)
occurs, and 7(0) is the time (y ear) of start of life o f the equipment.
As with TLTI, GDI 2 is not “equivalent'5to TEWI, unlike GDI I which is proportional to TEWI
provided GDI 1 and GWP have been evaluated over the same integration time. A further
similarity between TLTI and GDI 2, is that it represents the more immediate warming
contribution of high radiative forcing and short lived refrigerants. Whilst both GDI concepts
represent a more significant interpretation of the consequences of the emissions associated w ith
refrigerating equipment, they are considerably less reliable in terms o f representing the likely
impact due to the accumulated uncertainties associated with preceding evaluations upon which
GDI depends. In addition, an important consideration from the literature review was that the
consequence of “individual” emissions is largely dependent upon the conditions and
concentrations of gases at that time. In the case of greater preceding emissions, latter emissions of
a given quantity will result in more severe additional consequences. In other words, the impact of
increasing emissions is believed to cause exponentially worse climatic effects and associated
societal penalties. Thus, evaluation of GDI (and TLTI) on an individual basis may under
represent the sum o f all emissions.
81
SECTION 4: RESULTS AND DISCUSSION
4.1 Implementation of model

The model described in Section 2 was used to predict refrigerant and CO: emissions resulting
from the production, use and disposal of a refrigerating system, and the various consequences
measures detailed in Section 3 were used to understand the overall implication of the emissions.
To address the objectives of this study (Section 1.3), the model was used to examine the
following:
- breakdown and uncertainty of emission sources,
- sensitivity of transient influences,
- effects of system design characteristics on emissions,
- influence of material and refrigerant handling activities, and
- interpretation of consequence of emissions.
Consideration of the first four items should provide a better understanding of the influence of a
given parameter within a TEWI calculation. The second, third and fourth items are essentially an
in-depth investigation o f the uncertainties identified in the first case. The last item will help to
provide a broader overview of the consequences of emissions. In order to evaluate different
parameters, a reference system has been established and the selected input values (based on mean
values within the ranges identified unless stated otherwise) are provided in Table 14, which
represents a commercial refrigeration cabinet. Basic outputs for the system performance are also
included. The assessment of different parameters addresses variations of these reference inputs.
4.2 Breakdown and uncertainty of emissions

The model was used to generate data for the reference application detailed in Table 14, and for
each stage emissions w ere broken down into those from energy use and from refrigerant release
(converted into CO 2 equivalent using GWP(IOO)).
4.2.1 Breakdown of emission sources

Figure 16 show s a full breakdown of emissions from each element for SOL, In-use and EOL
stages. (Data for in-use stageâre presented for a one-month period only so that the data for other
stages is more visible.) The model was run for three different input settings, corresponding to
minimum, mean and maximum values of emissions, and these are indicated on the graph. These
settings included parameters for ageing effects, refrigerant handling behaviour, leak tightness and
recycling, all of w hich represent the input ranges for each sub-model as detailed in Section 2.
82
Table 14: Reference input parameters and basic outputs
Environment
External temperature Ambient Ambient temperature steps Monthly *
Design ambient 32°C Day/night ambient temp difference 2.3 K
Application
Design refrigerant R22 Application temperature -20°C
Refrigerant in use R22 Thermostat differential setting 3K
Application basic rating 250 W K'1 Application thermal mass 850 kJ K'1
Application constant rating 0.2 Operational life 8 years
Application heat load source Room Fan power constant 7 kJ m'3
System components
Evaporator diameter 0.009 m Condenser diameter 0.009 m
Total evaporator tube length 60 m Total condenser tube length 80 m
Number evaporator passes 6 Number condenser passes 4
Evaporator HTF flow rate 2.5 m V 1 Condenser HTF flow rate 4.5 m3 s'1
Evaporator int/ext area ratio 50 n r m 2 Condenser int/ext area ratio 60 n r in'2
Suction pipe diameter 0.02 m Discharge pipe diameter 0.015 m
Suction pipe length 4m Discharge pipe length 1m
Refrigerant handling ser\hcing
Number service visits per year 1 Recovery fraction 0.3
Number of top-ups per year 1 Recovery pressure 65 kPa
Pressure/leak test gas R22 Probability of venting when handling 0.2
Charging phase Liquid Frequency of handling per visit 0.3
Distance travelled per visit 75 km Material recycling None
Condenser cleaning No Oil and filter drier change No
Model selection
Evap/cond charge sensitivity index (n) 1.0 Annual leak rate increase 0%y1
Include evap/cond fouling Yes Include compressor wear Yes
Basic outputs (design conditions)
Refrigerant charge 2.49 kg Refrigerating capacity 14.6 kW
Annual leak rate (first year)1 8.1 % Compressor power 10.1 kW
Compressor swept volume 58.5 m3h_1 Coefficient of performance 1.47
Evaporating temperature 4.5°C Condensing temperature 48.3°C
t Ambient for London. UK (BS 5925, 1991). * Reference values in Table 16.
83
It was found that the dominant sources of emissions were primarily from operating energy
consumption during in-use stage, and then refrigerant leakage from the system and from
refrigerant handling during servicing. Energy consumption during SOL, refrigerant releases
during servicing and EOL were also significant. SOL refrigerant leakage and energy consumption
during servicing and EOL appeared small. Therefore, if a simplified approach was to be used,
then SOL refrigerant leakage and energy consumption during servicing and EOL should be
omitted.
Equivalent emissions (kgCCh)
o 5000 10000 15000 20000 25000
SOL, com pnt materials

SOL, com prt m anuf
SOL assenfcly
SOL leak te st
SOL transport
SGL, charging j
SOL reft dist i
SOL reft pro dn
In-use, recovery
0
In-use, evacuation
In -use, charging
In-use, venting
In-use. transport
In-use,reft (system)
In-use,energy (system)
BOL transport
ECL, recovery
BOL venting
EOL evacuation
BOL, inch (reft)
ECL, h c h (energy)
Figure 16: Breakdown o f emissions from energy and refrigerant fo r different stages (based on
GWP(100), In-use values for a single year only)
The breakdown of emission sources was extended over a range o f equipment sizes, and grouped
(according to life cycle stages) results are presented in Figure 17, and the values for the In-use
stage are for one year only. For the different sized equipment, model inputs were as Table 14,
although the application basic rating and thermal mass were changed proportionally, and system
component sizes were changed in similar proportions to maintain the same cycle COP as the
84
reference case. The data shows some variation in proportion to their emissions, particularly as the
capacity of equipment reduces. For example, with the larger capacity’ system, emissions from
operating energy consumption account for 85% of the total, whereas for the smaller capacity
system, it is about 55% o f the total. The two main sources that produce a greater proportion of
emissions for the smaller system are system leakage and refrigerant emissions from servicing.
Thus, it is pertinent to pay more attention to modelling refrigerant emissions for smaller
equipment (particularly to emissions during refrigerant handling), whereas more focus should he
on modelling energy7consumption in larger equipment.
i*
|
5* 0 . 1%
©
c
a
ao SOL, reft
In-use, reft oper
- D - SOL energy
- O - In-use. enxgyoper
CL
In-\ee, ret serv In-ise, energrserv
EOL, ret - O - EOL energy
0 10 20 30 40 50 60
Design cooing capacity (kW)
Figure 17: Variation o f contributions from different emission sources with size o f equipment
4.2.2 Uncertainties of emissions

The range of proportions o f each contribution is shown in Table 15, relative to the total emissions
for the minimum, mean and maximum values, also showing the extremities of the uncertainty.
Table 15: Range o f emission contributions from different sources
Minimum (relative to Mean (relative to mean Maximum (relative to

Element
max total) total) min total)
SOL. refrigerant 0.0% 0.5% 1.3%
SOL. energy’ 0.3% 2.3% 8.6%
In-use. system leak 2.5% 22.4% 58.1%
In-use, system energy 35.1% 70.0% 94.7%
In-use, service refrigerant 0.7% 3.4% 15.5%
In-use, service energy 0.0% 0.1% 0.1%
EOL, refrigerant 0.3% 1.2% 6.8%
EOL. energy 0.0% 0.1% 0.1%
Total - 100.0% -
85
In the example, the contribution of SOL varies between negligible (<0.5%) to about 10% when
the SOL emissions are high and the total of in-use emissions are relatively low. EOL emissions
cover a similar range. In-use emissions from energy consumption can vary significantly, from
35% to almost 95% of the total, and although always lower, refrigerant leakage can be from 2%
to over 50%. Including releases during servicing, total contribution o f refrigerant could account
for over 50% o f lifetime emissions. Overall, the total variation in TEWI for the given system and
environmental conditions ranges from -30% to +140% about the result based on the mean input
value for each selected parameter (not the average of the results). With a lower GWP refrigerant
(such as HC) this was found to be smaller, within the range -10% to +80%. This strongly
identifies the importance of accurate modelling of TEWI and detailed accounting of parameters
that influence in-use system leakage, energy consumption and refrigerant emissions from
handling. Similarly, the results of any TEWI evaluation should only be presented with the
expected uncertainty associated with the detail of input parameters.
200 -i—
ambient (27CC)
Varying mean
■ / monthly ambient
O 150 -
8 Constant mean
S
2 I / ambient (10.4CC)
' Varying monthly
j , and day/night
/ ambient
cr
v
50 -
c s <?cp
o o V
I o £ Q
Figure 18: Variation in monthly emissions from energy consumption fo r different ambient
scenarios
4.3 Sensitivity to transient influences

The impact of various parameters was assessed to help understand their influence, and suitability
of inclusion within the entire model. Three key time-dependent influences were considered:
- External conditions (ambient temperature variations)
- Ageing effects (fouling, compressor wear and increasing leak rate)
- Power station emissions
Each of these was addressed individually, and then in combination.
86
4.3.1 External conditions
Figure 18 shows the differences between monthly energy consumption of the compressor for
different assumptions: monthly and dav/night variations, average monthly variations and a
constant ambient. The results show a significant difference between each of these three scenarios,
where the total annual emissions from energy' consumption with a constant ambient (assumed
27°C) is greatest (2340 kgCO:), and by varying monthly ambient temperature (1190 kgCC>2 ) and
including changes for day/night temperatures (950 kgC 02) produces a notable reduction in
emissions. As an alternative, the model was tested with an average annual ambient temperature of
10.4°C, which produced an annual energy' consumption o f 800 kgCCK This value is significantly
low er than the varied monthly/dav/night scenario, confirming that using a constant ambient
approach will always be erroneous, and that modelling with incremental temperatures throughout
the year is valuable. When the heat load and heat rejection temperatures have a stronger
dependence on ambient variations, the errors of using an assumed ambient will be even greater.
36000
c
u X—*
.2?
K 32000
o
(A
CC
£4>
28000
.JU
c —o—constant
> —0—comprwear
5U"
24000 —o—evap/cond folding
"3 —6 - dRdt = 6 y-1
1 - x - a ll
< 20000
2004 2006 2008 2010 2012 2014
Year
Figure 19: Influence i f various ageing effects on emissions from system energy consumption
4.3.2 Ageing
The influence of heat exchanger fouling, compressor w ear and refrigerant leakage on annual
emissions from energy' consumption is shown in Figure 19. Individually, compressor wear has the
smallest impact, but rises almost exponentially with time. This occurs because of the feedback
effect of the resulting lower system capacity causing longer running times and therefore more
wear. The effect of fouling is seen to have a significant influence, although the degree of fouling
is also largely dependent upon the HTF (i.e., w ater, air, etc.) and local contamination of the HTF
(e.g., hard or soft w ater, salty air, etc.). Typical rate of increasing leakage rate also has a notable
effect, but this will be largely' dependent upon the sensitivity7of the system to leakage (as
87
discussed later). Both fouling and increasing leakage are seen to produce an increase of annual
emissions at a slower rate.
In combination, the influence of compressor wear, fouling and increasing leakage w as found to
produce emissions about 10 % higher in the final year o f the equipment life compared to the first
year. However, it is also observed that the total increase is less than the sum of the individual
contributions, when it would be expected that the contributions would at least be cumulative. This
is explained when considering the performance o f the system itself. The application temperatures
are within the required range when the individual transient influences are accounted for. When all
three influences are included the capacity of the system degrades to a point where the required
application temperature cannot be achieved, and with a smaller temperature lift, the impact of
greater energy consumption is dampened.
28000 -
£
E
24000 -
-O —constart
-O —annual change
- A —morthly change
-O —dayvryît change
20000
2004 2006 2008 2010 2012 2014
year
Figure 20: Influence o f change in power station emissions fo r electricity production
4.3.3 Power station emissions

The effects of changes in C 0 2 emissions from energy production are shown in Figure 2 0 . With
the variable emission factor neglected, the system energy consumption produced constant
emissions for each year. When applying anticipated changes in emissions factor, the annual
emissions reduce and would be expected from the data in Section 2.3.2. Accounting for annual
changes in emissions factor (£ ) only (using only the first term in equation 23), it is seen that
there is a constant reduction in emissions over the equipment life, where the final emissions are
about 6 % lower then the first year, and the total lifetime emissions are reduced by about 4%. If
monthly variations in emissions factor are introduced (include the second term in equation 23),
88
the same trend is observed, but the emissions for the starting years are higher by about 3%. This
is due to the higher emissions factor occurring at the same as the higher power demand of the
system, i.e., during summer months. Including additional day/night variations in emissions factor
(full equation 23), the starting emissions are greater still at (5%), because daytime emissions are
greater which again correspond to higher daytime power demand. In this example, the difference
between detailed accounting for time-dependent emissions factor, and constant emissions factor
on total lifetime emissions is negligible. However, this may not be the situation if different load
profiles exist, and particularly in the case of heat pump applications.
4.4 Influences of equipment design

It is important to understand how the design of equipment influences GHG emissions, particularly
since certain design parameters may lead to lower emissions from one source, but cause more
emissions elsewhere. Where trends in emission reduction through certain design parameters are
found, these can be used as a basis for improved environmental design of new equipment. The
size and type of several components were changed:
- Circuit piping and connections
- Heat exchanger sizes
- Compressor size
- Refrigerant
This will provide additional guidance on how to minimise emissions through design options.
4.4.1 Pipework and connections

The choice of components and connection types used in circuit piping was found to have a
notable change on both refrigerant emissions and emissions from energy use, as seen in Table 16
for the five different cases, including the reference systems as Case 1. In Case 2, flare
connections are changed to brazed connections. Case 3 is with the service valve removed and
swaps the TEV for a capillary tube, and Case 4 has no solenoid valve, or controls (e.g., pressure
switch and connecting line). Case 0 represents the w;orse case, where all pipe connections are
flare joints.
Compared to the reference Case 1 , equivalent emissions for Case 2 from refrigerant were reduced
by 16%, but energy consumption dropped by only 1 %, although the mean application
temperature is about 0.5 K lower. With Case 3, an additional reduction of about 2% is observed,
although it should be noted that an absence o f service valve could result in an increase in
refrigerant emissions because of additional handling wfren cutting into the system. Further
89
emissions from energy consumption are negligible, although there is a slight improvement of
application temperature. Case 4 shows an additional 4% reduction in refrigerant emissions, and
since it is the situation with minimal components and virtually no mechanical connections, it is
considered the best overall reduction in emissions. Again, there are no more emission reductions
from energy consumption, but significant benefit is seen in lower application temperature. Where
leakage is about twice the rate as the reference case, emissions for Case 0 are 31% for the
refrigerant and 3% higher for energy consumption. Consistent with previous trends, application
temperature is almost IK higher.
Table 16: Effect o f refrigerant leakage from different component assemblies
Case 1
Component Case 0 Case 2 Case 3 Case 4
(Reference)
Coil 2 2 2 2 2
Pipew ork/brazed joint 0 12 18 20 20
Service valves 1 1 1 0 0
Solenoid valves 1 1 1 1 0
Expansion valves 1 1 1 0 0
Control lines 2 2 2 2 0
Control equipment 2 2 2 2 0
Filter 1 1 1 1 1
Compressors 1 1 1 1 1
Flare joint 18 6 0 0 0
Annual leakage rate 17.9% 8.1% 3.2% 2.8% 1.6%
Refrigerant contribution (kgC02) 13165 10067 8486 8373 7978
Energy contribution (kgC02) 252446 250468 247759 247652 247228
Total (kgC02) 265612 260536 256245 256025 255206
Mean application temperature (°C) -20.49 -21.16 -21.46 -21.51 -21.68
4.4.2 Heat exchanger size

Figure 21 show's the refrigerant and energy use contribution to GHG emissions when the
exchangers are between 50% and 200% o f their original ‘‘size”. The size o f both condenser and
evaporator was changed in two ways: (a) vary tube length, whilst maintaining the same number
of circuits and the same HTF flow, and (b) vary tube length and HTF flow7 proportionally.
90
30000 -I----
O
Si 20000
10000 - total, const V_HTF

- eneigy. const V H T F
- total var V_HTF
ena-Q. var V H T F
0% 50*. 100*/. 150*/o 300°. 250%

Relative evaporator and condenser see
Figure 21: Effect o f evaporator and condenser size on annual emissions
With a smaller heat exchanger the corresponding refrigerant charge is smaller and therefore the
amount o f refrigerant emitted is less (given the same leak rate), so the refrigerant contribution is
less. Where the HTF flow rate is reduced with exchange size, the emissions are less because the
benefit in lower fan energy consumption outweighs the higher heat transfer coefficient associated
with maintaining the higher flow rate. However, where the HTF flow7 rate increases with
exchanger size, the additional power consumption offsets the improvement in heat transfer and
total emissions increase. Where HTF flow remains constant with larger exchanger size, the rate
o f improvement in emissions becomes less, and it can be seen that a further enlargement of the
exchanger would begin to increase emissions due to the refrigerant emissions. Obviously, a higher
leak rate would produce this reversal in the trend sooner. These observations imply that the size
of the condenser and evaporator can be “optimised"’ when considering the combination of increase
o f refrigerant emission and decrease in emissions from energy7 consumption.
40000
E 20000
=r 10000
Figure 22: Influence o f system sensitivity to charge size on annual emissions from energy
consumption
91
Additional emissions due to manufacture of larger components is dwarfed by the savings from
lower compressor energy requirement. Changing the size of the exchanger equates to about 20%
of the production emissions, but only accounts for less than 1 % o f the total emissions.
Another aspect associated with exchanger design that influences emissions is the sensitivity of the
system to sub-optimal charges. Condensers and evaporators can be designed so that the system
performance becomes less sensitive to a reduction in charge size due to leakage. Figure 22
presents results for annual emissions from a system with three different leak rates corresponding
to three of the cases detailed in Table 16, where different sensitivity indexes w ere used.
3000
application
temperature w ithin
required range
- 2000
Case 4, n - 13 —o—Case 0. n=l/3o

Case 4, n=l 17 —a—Case 0. n=l/17
■Case 4, n=l 30 - o —Case 0. n=l/30
1000
-25 -20 -15 -10 -5 0
Mean monthly application tenperature (°C)
Figure 23: Effect o f sensitivity to charge size on system performance
For n = 1/3, the trend is as would be expected: emissions from energy consumption reduce as
leak rate reduces because the system is closer to its optimum charge. However, where n = 1/17
and n - 1/30, the emissions are actually lower for the high leak rate (Case 0). With reference to
the earlier discussion on ageing effects, the same phenomenon is occurring here, in that these
conditions lead to significantly reduced capacity, so power consumption is less. This is explained
further with Figure 23, where the mean monthly application temperature and corresponding
equivalent emissions is plotted for two different leakage conditions (Cases 0 and 4). Where n =
1/3, the application temperature is within the required range, and emissions from energy
consumption vary with the monthly ambient. However, where the system sensitivity to charge size
increases (n = 1/30), the emissions reduce as in Figure 22, but also the mean application
temperature rises away from its set point, and thus fails to serve its purpose. This effect is more
pronounced for the higher leak rate (Case 0). These observations demonstrate the importance of
good heat exchanger sizing/design.
4.4.3 Compressor size
Compressor swept volume was varied by setting the design ambient to a different temperature,
which gives a larger design heat load and higher condensing temperature. Results are shown in
Figure 24 where the design ambient is indicated for the corresponding size of the compressor.
Changes in refrigerant emissions increase slightly because o f releases of residual refrigerant
within the compressor oil, and similarly, material production emissions increase a little, but all
are relatively small.
The main reasons for the differences in emissions seen in Figure 24 are due to changes in energy
consumption. It is seen that using a compressor with smaller swept volume produces lower
emissions. This is because the system balance temperatures are closer to the HTF temperatures,
so the compression ratio is less and therefore compressor COP is higher. In addition, a lower
refrigerant flow rate reduces pressure drop around the system, again leading to smaller
compression ratio, and the lower number of compressor starts also means that less energy is
wasted. However, it should be noted that with compressor sizes below 80%, there is a rapid
reduction in emissions. As with earlier evaluations, this was found to be partly due to the system
possessing insufficient capacity to achieve the required application temperature.
40000
20000 -
-1 1 .r c
0°/. 40% 80% 120% 160% 200%

Relative compressor size
Figure 24: Effect o f compressor size on emissions from system operation
4.4.4 Refrigerant selection

The final assessment o f design issues relates to the refrigerant selection, which is particularly
important since fluid choice is perhaps the main driver for the popularisation of the TEWI
concept. As identified by the Author (Colboume and Suen, 2000), the subject of comparative
93
refrigerant performance is complex and controversial and for this reason, only a limited
assessment will be carried out here. In addition, the evaluations were simplified by only
considering pure refrigerants, so that the complexities associated with composition shift in
zeotropes are avoided. Using a similar principle to the cost-based approach described by Douglas
et al (1999), the influence of selected refrigerants are based on three different basis for
comparison:
a) Fixed system design. Only the refrigerant selection was changed, thus representing different
fluids being used in the same system. This could be interpreted as a retrofit type comparison
(Table 17).
b) Fixed system capacity at design conditions. Compressor size was changed so the design
capacity is fixed for all refrigerants at design conditions, but the rest of the system remains as
(a) (Table 18).
c) Fixed system capacity and COP at design conditions. Compressor size was changed as (b),
but although compressor COP is the same for all refrigerants, evaporator and condenser size
and HTF flow rate (to maintain velocity) were adjusted equally to maintain the system COP
at design conditions (Table 19).
Table 17: Re frigerant comparison (a) with fixed system design
Parameter R152a R134a R290 R22(ref) R1270 R143a R125

Relative exchanger size (%) 100% 100% 100% 100% 100% 100% 100%
Comp. swept volume (m3 h'1) 58.5 58.5 58.5 58.5 58.5 58.5 58.5
COP at design (-) - - - 1.47 - - -
Mean COP (-) 2.44 2.39 2.34 2.15 2.33 2.16 2.04
Mean system capacity (kW) 11.79 11.93 15.53 16.59 20.21 19.16 18.81
Mean Tapp <°C) -15.18 -15.75 -21.39 -21.70 -21.53 -21.66 -21.61
Time Tapp = set point (%) 0% 8% 100% 100% 100% 100% 100%
SOL/ Refrigerant (kgCCb) 122 1839 4 2526 3 5524 5753

EOL Energy (kgC02) 3032 3110 3017 3634 3017 3233 3397
In- Refrigerant (kgC02) 360 5458 12 7675 10 16938 17756

Use Energy (kgCOo) 202880 207690 223581 238351 223958 236968 248654
Sub Refrigerant (kgC02) 482 7296 16 10201 14 22461 23509

total Energy (kgC02) 205912 210800 226598 241985 226975 240201 252051
Total (kgCCL) 206394 218096 226614 252186 226989 262662 275561
94
Each table shows variations of heat exchanger size (both evaporator and condenser), compressor
swept volume, COP at design, mean COP, mean application temperature, mean refrigerating
capacity’, and the proportion of time (over the year) that the application achieves design
temperature. Mean COP and temperature are time-weighted according to operating conditions.
Equivalent emissions for refrigerant and energy consumption are presented for SOL/EOL and In-
use stages. Seven refrigerants are included, and are listed in order of decreasing normal boiling
point (NBP). The reference refrigerant was R22, which means that the input parameters for
comparison (b) and comparison (c) were varied to match the original R22 design. Lastly, it is
noted that for each combination - whether the system is designed for a given refrigerant or it is
retrofitted - the charge size is optimum and it is assumed that superheat is always adjusted to the
prescribed value.
Table 17 shows that the mean COP and capacity approximately follow the NBP of the
refrigerant. Lower COPs and higher capacities occur with lower NBP refrigerant, which leads to
the higher emissions from energy consumption. This is due to the greater volumetric refrigerating
effect of low er NBP refrigerants. How ever, because the refrigerating capacity’ of the higher NBP
refrigerants is lower, the R152a and R134a systems were rarely able to achieve the average
application temperature set point, and consistent with earlier evaluations, this is reflected in the
loyv In-use stage emissions from energy-. Equivalent emissions from refrigerant release both for
SOL/EOL and In-use stage closely follows the GWP of the fluid, yvhereas emissions from energy
consumption at SOL/EOL stages are virtually equal for all systems. The total lifetime emissions
are shoyvn relative to the R22 system in Figure 25. Considering the trends in emissions from in-
use stage energy’ consumption only, it is evident that refrigerant GWP can have a major influence
on total emissions as can be seen by the significantly loyver values for R290 and R1270, and the
maximum emissions for R143a and R125.
For comparison (b), the design evaporating capacity was maintained at a design application load
of 14.6 kW. Note that all systems now consistently meet the design application temperature. The
consequence of matching compressor size to the heat load for different refrigerants is that high
NBP refrigerants suffer degradation of COP at design conditions, although mean annual COP is
relatively constant. This is because (for otherwise fixed conditions) increasing condensing
temperature degrades cycle efficiency as it approached the critical point. Typically, high NBP
refrigerants have low?er critical temperatures so raising the condensing temperature means that a
higher NBP refrigerant approaches critical temperature more rapidly than the lower NBP
refrigerant, and therefore efficiency decays at a higher rate. Since their condensing temperature at
95
design conditions is intrinsically higher than under usual ambient conditions, the relative
difference between cycle efficiencies of the different refrigerants is amplified, whilst under usual
ambient conditions this influence is less pronounced.
Table 18: Refrigerant comparison (b) with fixed system capacity at design conditions

Comp swept volume (m3 h'1) 124.2 143.3 67.9 58.5 50.8 59.9 75.7
COP at design (-) 1.713 1.572 1.668 1.47 1.573 1.326 1.09
Mean COP (-) 2.14 2.06 2.30 2.15 2.33 2.17 2.04
Mean Tapp ("C) -21.70 -21.72 -21.72 -21.70 -21.69 -21.54 -21.52
SOL/ Refrigerant (kgCO?) 124 1874 4 2526 3 5522 5787

EOL Energy (kgCQ?) 3798 4066 3140 3634 2911 3251 3620
In- Refrigerant (kgCO?) 354 5414 12 7675 10 16942 17813

Use Energy (kgC02) 241502 250416 226289 238351 223091 236230 259762
Sub Refrigerant (kgC02) 478 7287 16 10201 14 22463 23600

total Energy (kgC02) 245300 254482 229429 241985 226002 239481 263382
Total (kgCOz) 245778 261770 229445 252186 226016 261944 286982
However, emissions from energy consumption are closely correlated to mean COP, but bear no
relationship to the COP at design conditions. The differences in amount of system materials due
to changes in compressor size are seen to have only a negligible impact on overall emissions. As
with (a), equivalent emissions from the refrigerant vary considerably with GWP. Comparison of
total emissions in Figure 25 shows results that are more favourable for R290 and R1270, and all
lower NBP refrigerants have total emissions that remain similar to (a).
Finally, for the comparison (c), the capacity was kept constant and the COP was maintained at
1.47 under design conditions, thus necessitating changes in both exchanger lengths and HTF
volume flow rate. Despite this, it is seen that the mean COP still varies significantly between
refrigerants because of their individual behaviour at off-design temperatures, and in fact, the
mean COP clearly follows the refrigerant NBP. However, there no longer appears to be a
correlation betw een mean COP and emissions from energy production and this a result o f the
neutralising effect that the larger mean capacity’ of low’ NBP refrigerant has on compressor
power. Although the capacity matches at design conditions, the system using low NBP
96
refrigerants becomes significantly oversized at the more usual off design conditions, which leads
to the overall system efficiency suffering, and losses due to more frequent cycling. A greater
proportion of the contribution of energy consumption for R143a and R125 is due to greater fan
power because of larger heat exchangers, and similarly a lower contribution from fan power is
evident for the higher NBP refrigerants. It is also noted that R143a and R125 required the
additional airflow in order to achieve equal COP at design conditions.
Table 19: Re frigerant comparison (c) with fixed system COP at design conditions

Comp swept volume (m3 h 1) 166.1 153.2 78.7 58.5 54.2 52.7 49.8
COP at design (-) 1.47 1.47 1.47 1.47 1.47 1.47 1.47
Mean COP (-) 1.94 1.96 2.08 2.15 2.19 2.31 2.45
Meai. Tapp CO -21.35 -21.67 -21.72 -21.70 -21.66 -21.68 -21.53
SOL/ Refrigerant (kgCO?) 91 1519 3 2526 3 6491 10807

EOL Energy (kgC02) 3762 3859 2988 3634 2803 3426 4685
In- Refrigerant (kgCO?) 226 4077 10 7675 9 20606 36694

Use Energy (kgCO?) 238933 245586 231023 238351 225952 236422 267801
Sub Refrigerant (kgCO?) 317 5595 13 10201 12 27097 47502

total Energy (kgCO?) 242695 249445 234011 241985 228755 239849 272485
Total (kgC02) 243011 255040 234024 252186 228766 266946 319987
The three comparisons detailed here reveal some key observations. First, it is seen from the
analysis that the equivalent emissions can van,' significantly depending upon refrigerant selection
and system design. It is important to match system and refrigerant in order to assure the required
application temperatures. More importantly, the sizing of components should really be done at
conditions closer to the average temperatures during normal operation as it is clear that
optimising for one ambient can “reverse’’ the optimisation at a different temperature. To
demonstrate this, the emissions were calculated for a design ambient of 11°C (which is near the
average temperature over the year) whilst maintaining the same capacity- and COP of R22 as in
comparison (c), and the results are also shown in Figure 25.
It is seen that the results are somewhat contrary to previous comparisons, where the lower NBP
refrigerants have low-er emissions than those with high NBP. Further, the emissions for R152a
97
and R134a are highest in this case that for previous comparisons, whereas the other refrigerants
produce lower emissions, including the reference R22. However, it was found from the analysis
that this approach (matching capacity and COP at an appropriate ambient) can cause the system
to be incapable of achieving the required application temperature at certain times. Also
noteworthy is that the mean annual COP for all refrigerants was virtually identical to the design
COP; similarly, mean annual capacity was close to the design capacity. Although the mean COP
and capacity were nearly identical for all refrigerants, the emissions from energy production for
system operation varied significantly with then same ranking as total emissions, where R152a and
R134a were 10% and 20% higher than R22 respectively, R290 and R1270 were 5% lower and
the others were close to R22. This variation is explained by the additional operating duration for
lower pressure refrigerants at the higher ambient temperatures. Finally, any changes in system
design to reduce emissions through capacity matching or efficiency enhancement will not be
offset by consumption of additional materials.
1.4
■ comparison (a)
B comparison (b)
□ conparison(c)
0 11°C design arrfcienl
R152a R134a R290 R22 R1270 R143a R125
Figure 25: Comparison o f total emissions fo r different refrigerants
In terms of refrigerant selection, a general observation is that R290 and R1270 always result in
lowest emissions when the required application temperature is achieved. This is partly due to
favourable thermal properties, w hich provide good efficiency, and partly because of their low
GWP compared to the HFCs.
4.5 Effects of material/refrigerant handling and operations

As with equipment design, it is important to understand the effect that certain material handling
and other human operations have on GHG emissions, as a particular measure can lead to lower
98
emissions from one source, but cause more emissions elsewhere. In order to evaluate the influence
of handling and other operations on emissions, the following cases were addressed:
- Poor and best practice during refrigerant handling, and recycling
- Frequency of condenser and system cleaning
- Frequency of checking and servicing leaks
Understanding this process will provide useful insight into optimisation of operational activities
to reduce emissions.
15000
□ best practice
□ normal
Bnannal'recjcfed
10000 B worse practice
Figure 26: Comparison o f emissions fo r different service and handling practices
4.5.1 Refrigerant handling practices and recycling

Best practice, normal and poor practice for refrigerant handling w as characterised by setting the
model parameters according to Table 20. Best practice accounts for the refrigerant being used in
a manner that aims to minimise emissions. Normal practice is what the author considered as
typical for the companies involved, and poor practice represents the case where refrigerant
containment is virtually neglected.
The reference system in Table 14 was used to generate emissions for SOL, In-use (for one year),
and EOL, and input parameters adjusted according to Table 20. The results are presented in
Figure 26. Compared to normal practice, the results for best practice show' significant reduction
at SOL, in-use servicing and EOL refrigerant handling. Overall, there is a 5% reduction over a
10-year lifetime, although in terms of refrigerant handling only, this increases to 40% reduction.
Conversely, poor practice increases total lifetime emissions by almost 20% and in terms of
99
refrigerant emissions only, 200%. Where a refrigerant with low GWP is employed then there
would be negligible effect caused by handling practices.
Table 20: Input parameters fo r refrigerant handling practice
Parameter Poor practice Normal practice Best practice

Recovery pressure 100 kPa 65 kPa 30 kPa
Frequency of refrigerant handling per visit 0.3 0.3 0.3
Probability of venting during handling 0.9 0.2 0.1
EOL recovery fraction 0.1 0.3 0.9
Refrigerant
Leak test gas Refrigerant Inert gas
tracer
Charging phase Liquid Liquid Vapour
Lastly, Figure 26 also presents results for normal practice where materials and refrigerant have
been recycled. It is seen that recycling only affects SOL energy emissions, reducing them by
about 50%. However, over the total lifetime this accounts for 1-2% of emissions. If a refrigerant
with low GWP were used, these differences would be halved since the contribution from
refrigerant production emissions at SOL would be minor in the first place. There was also a
difference at EOL energy' emissions because there is less for incineration, but this again is
negligible.
□ energy’ ■ total
op
0c
B
1J3
C
.a
no ageing ageing ageing, 1 ageiqg, 1 ageng, 1 agerg, 2

only a n d clean sys cfean sys clean, sys clean,
1 cond 2 cond
cfean cfean
Figure 27: Influence o f condenser cleaning and system cleaning on emissions
100
4.5.2 Condenser and system cleaning
Section 4.3.2 provided results for variation in emissions from the system due to the influence of
exchanger fouling and compressor wear. Although both of these cause degradation of system
performance and therefore increased emissions, their impact can also be reduced by removing
fouling from exchanger surfaces by cleaning, and compressor wear can be lessened by ensuring a
clean internal system, by replacing the filter/drier and compressor oil. To quantify these
processes, the model was run for different frequencies of maintenance visits where the
contribution of fouling is halved whenever the condenser is cleaned and the reduction in
volumetric efficiency is reduced proportionally to the number of oil/filter/drier changes. For
equation (43) and (44), substitute the constant a for a /(1 + N sen,) and for equation (45) assume
ix l k )f0ui = 0.5 • ( x / k) foul for every visit.
ida
gein
g ag eing ageng, 1 agekjj. 1 ageng.1 a geip.2
|only| caidcbansjscban sys c lean
, sys clean
,
H H 1co nd 2co nd
H cban cban
Figure 28: Influence o f condenser and system cleaning on mean application temperature
Figure 27 shows the resultant annual emissions for both energy contribution and total emissions
after 8 years of operation, relative to the case for no ageing (i.e., zero years). Data is provided for
the case of no cleaning, cleaning the condenser once per year, cleaning the internal system once
per year, and for both practices once and twice per year.
As shown here and in Figure 19, the influence of ageing causes a notable increase in emissions.
Both condenser and internal system cleaning reduce this effect slightly in terms of emissions from
energy use, although total emissions increase above the case for no cleaning because breaking
into the system results in increased refrigerant release. Cleaning the system twice in a year
produces a minor improvement, but still the effect of additional refrigerant emissions can
outw eigh the benefits of reduced energy use from minimising compressor wear. Note that if the
101
refrigerant possesses a small GWP (such as an HC) then there would be value in both condenser
and internal system cleaning. The influence of additional emissions from travelling to the site
more frequently is negligible. This is seen in Figure 28 also, where the process of maintaining
good performance allows the mean application temperature to be closer to the required design
temperature.
25000
■ refrissrant (operation) □ erenrv (operation)
□ refrgerart (service) total

5 20000
Si
e
15000
11 10000
3
5000
1 serv<'l 1 serv/2 1 serv/3 2 servs-2 2 servs3 3 seivs/3

top-up top-ups top-ups top-ups top-ups top-ups
Figure 29: Effect o f servicing repair and top-up frequency on emissions
4.5.3 Leak checks and repair

Loss of refrigerant due to leakage is accounted for in the model by reducing exchanger surface
area available for heat transfer in proportion to reduction in charge level. The influence of service
repair and top-up in terms of sub-optimal charge effect on emissions is examined here. Six
scenarios are considered, where frequency of service for repairing the leak, and topping-up the
charge with or without repair are varied. Only topping-up the system means that the system
recovers its optimum charge size but the annual leak rate remains the same. Repairing the leak
and topping up reduces leak rate and regains optimum charge.
Figure 29 show's the annual equivalent emissions of refrigerant during operation and servicing,
from energy consumption and the annual total. In terms of refrigerant emissions, those from
operation reduce with more frequent repair. However, a larger number of top-ups increase
refrigerant emissions because of releases during refrigerant handling, although this increases
when repairs are made also as the refrigerant must be recovered from the system. The emissions
resulting from the service person transportation are very small when compared to the equivalent
emissions from refrigerant handling during servicing. In terms of emissions from energy
consumption, there is a general increase, which seems contradictory since the efficiency of the
system will be improved. If mean application temperature is considered, then this observation can
be explained. Figure 30 shows that more frequent servicing and top-ups lead to improved
application temperature, which is away from the design value because the loss of charge also
causes a reduction in refrigerating capacity. Thus, whilst emissions from energy consumption
increase, the purpose of the equipment is better achieved.
Figure 30: Effect o f servicing repair and top-up frequency on mean application temperature
(design application: -20°C)
From this discussion, it appears that there is a balance between better performance o f the
equipment, additional refrigerant emissions from handling and whether system energy
consumption will be reduced or increase through less system leakage. Thus, operation procedures
can be optimised to achieve lowest GHG emissions. These results are strongly dependent upon
the sensitivity of the system to reduced charge and the behaviour of the service person when
handling refrigerant. In particular, if a refrigerant with negligible GWP were used the situation
would be different from the case above because the benefits of serv icing and topping-up on
energy consumption will never be offset by the negative impact o f additional refrigerant
emissions.
4.6 Comparison with BRA method

Finally, a comparison was made against the only formalised TEWI methodology currently
available, which is published by the BRA (1996). The standard methodology w-as followed, and
all data used in the calculation was taken from the document. All other application data was taken
from the reference system data in Table 14. Table 21 reproduces the format for calculating
energy consumption, where the power drawn was obtained from running the system model
103
(without ageing influences) for the specified ambient temperature. This calculated power
consumption data, was then fed into the standard TEWI calculation table from BRA, which is
reproduced in Table 22, alongside data generated using the model in this study.
Table 21: Energy calculations according to BRA method
Run time per Energy

Segment name Utility factor Ambient Power drawn
year consumed
Summer load - peak N/a 27°C 9.11 kW 116 h 1057 kWh
Summer load - base N/a 22°C 8.71 kW 632 h 5505 kWh
Autumn/spring load - peak N/a 17°C 8.26 kW 2154 h 17792 kWh
Autumn/spring load - base N/a 12°C 8.25 kW 2834 h 23381 kWh
Winter load - peak N/a 7°C 8.08 kW 2234 h 18051 kWh
Winter load - base N/a 2°C 7.96 kW 788 h 6272 kWh
Total (compressor) 72057 kWh
Fans - condenser 100% N/a 1.31 kW 8760 h 6570 kWh
Fans - evaporator 100% N/a 0.75 kW 8760 h 11738 kWh
Total (ancillary ) 18308 kWh
Comparison betw een the results of the BRA method and the model in the present study show that
BRA produces significantly larger overall TEWI, where even the model calculations using the
maximum settings were 20% lower. For the refrigerant contribution, the modelled values are
between one half and seventeen times the BRA value, as reflected in the in-use and the retirement
or EOL emission values. This difference is attributed to the inclusion o f the numerous peripheral
release sources that the model accounts for, and the range in annual leak rates. The modelled
energy consumption is 25% to 40% lower than that estimated by the BRA method. The main
reason for this is that the model includes for compressor cycling and reduction in application heat
load in response to variations in ambient conditions, despite including additional emissions due to
SOL, servicing and EOL, and a reduction of power station emissions over time. Also of interest is
the proportion of total emissions due to refrigerant release: the BRA method gives less than 1%
whereas the model gives up to 10%. This comparison suggests that to provide a more accurate
representation of TEWI, the BRA method should at least include detailed system energy use
calculations, and improved assessment of refrigerant releases such as during refrigerant handling.
104
Table 22: Standard calculations according to BRA method
Values for BRA Values for present study

Input parameter
(1996) Minimum Mean Maximum
Refrigerant name R22 R22 R22 R22
Refrigerant charge size 2 .49 kg 2.49 kg 2.49 kg 2.49 kg
Annual Compressors 72057 kWh 17977 kWh 23252 kWh 29053 kWh
energy Ancillary 18308 kWh 1259 kWh 1628 kWh 2034 kWh
System operating lifetime 8 years 8 years 8 years 8 years
Refrigerant GWP 1700 kgCO: 1700 kgC02 1700 kgC02 1700 kgC02
LI (annual leak rate) 0.05 0.022 0.05 0.10
L2 (annual purge factor) 0 0 0 0
Leakage
factors SI (annual serv release) 0.0025 0.026 0.10 0.3
S2 (catastrophic failure) 0 0 0 0
Recovery7efficiency7 0.95 1.0 0.7 0
Electricity C 02 emission factor 0.53 kgC02 See 2.3.2 See 2.3.2 See 2.3.2
Direct Reft loss (operation) 1.05 kg 0.55 kg 3.93 kg 14.21 kg

effect Reft loss (retirement) 0.12 kg 0.04 kg 1.04 kg 4.92 kg
Total refrigerant loss 1.17 kg 0.59 kg 4.97 kg 19.13 kg
Refrigerant loss C 02 equivalent 1990 kgC02 1068 kgC02 9229 kgC02 34208 kgC02
Indirect emissions C02 equivalent 383150 kgC02 242573 kgC02 274909 kgCO? 282495 kgC02
Total equivalent warming impact 385140 kgC02 243642 kgC02 284138 kgC02 316703 kgC02
4.7 Interpretation of global warming

Section 3 detailed alternative measures for evaluating climate change impact, rather than
depending upon the GWP concept. The values of TLTI, GDI 1 and GDI2 were obtained from the
emissions model, and compared against results for TEWI. Calculations were made for a reference
system, and four other systems with differing characteristics: ( 1 ) high efficiency and high leakage,
(2) high efficiency and low leakage, (3) low efficiency and high leakage, and (4) low efficiency
and low leakage. These were adopted to help illustrate the differences in global warming impact
according to different measures when a refrigerant with a high F and r (R22) is compared
against a low F and r refrigerant (R290).
A first example is shown in Figure 31 where a new reference system (Table 14) is calculated for
R22 and R290 to show7 the difference between a high and low climate impact refrigerant. The
contribution from energy consumption and direct effect of refrigerant are separated for the R22
system, but only the total has been shown for R290. It is seen that the temperature rise is rapid
105
soon after the start of the system installation and within 15 years the maximum temperature rise,
or TLTI, is reached. Although the leak rate was Roper = 12%, the contribution of the R22
emissions accounts for over one-third of the temperature rise for the duration of the system life.
Following this peak, the impact on temperature drops rapidly and within 70 years all influence
from R22 emissions have ceased. However, the contribution from energy consumption has only
decayed by about one-quarter of its maximum. By comparison. AGTP of the refrigerant for the
R290 system is negligible and TLTI of that system is therefore a function of energy consumption
only.
35 T
-o -E n erg y (R22)
TLTI
30 - - f t- Refrigerant (R22)
- o - Total (R22)
GDI2 Total (R290)
O
X
&
£o GDI2
<
G D I1
2000 2020 2040 2060 2080 2100

Year
Figure 31: Comparison ofAGTP-ti me profile fo r high and low climate impact refrigerant
Table 23 presents full results for TEWI, TLTI, GDI1 and GDI2 for the reference system with
each refrigerant. Using GWP(IOO), TEWI of the R22 system is 10,653 kgCCL in comparison to
6,867 kgCCL for R290, being 36 % lower. This is a close comparison to TLTI, which indicates
the R290 system to be 39% lower than the R22 system. For the global damage indices, the R22
system GDI1 is 33% higher than GDI1 when R290 is used. Most significant is the difference in
GDI2, where the R290 GDI2 is 64% less than the R22 value.
The comparison between the different measures was extended by evaluating four different
systems, each representing different combinations of efficiency and leakage. The model input
conditions are based on a reference system, although for the high efficiency system, heat
exchanger sizes w'ere doubled, and half size for the low7efficiency system. In addition, a low7
leakage system is assumed to have annual Roper =5% , whereas the high leakage system has
annual Roper = 20%. The data in Table 23 show s a broad variation of values for each climate
106
change impact. As expected, the highest value of TEWI is for the low efficiency, high leakage
system, and the lowest value of TEWI are for the high efficiency, low leakage system. When
comparing data for the other measures, the sign is the same.
Table 23: Comparison o f climate impact fo r different systems
TEWI
System characteristics TLTI (K) GDI1 (K3~y) GDI2 ( K y 1)
(kgC02)
Reference Reference R22 10653 3.09 x 1019 8.46 x 10‘27 2.38 x 10‘17
efficiency leakage R290 6867 1.82 x 1019 5.63 x 10-2? 8.67 x 10'18
High R22 10341 3.06 x 10'19 8.94 x 10'27 2.18 x 10'17

leakage R290 7559 2.01 x 10'19 6.50 x 10'27 9.55 x 10'18
High
efficiency R22 9363 2.64 x 1019 7.51 x 10'27 1.76 x 10'17
Low’
leakage R290 7039 1.87 x 10'19 5.84 x 10‘27 8.89 x 10'18
High R22 12734 3.90 x 1019 1.07 x 10'26 3.00 x 1017

leakage R290 7124 1.89 x 1019 5.94 x 10 - 9.00 x 10’18
Low
efficiency R22 11478 3.19 x 1019 8.64 x 10-27 2.45 x 1017
Low
leakage R290 6904 1.83 x 10’19 5.67 x 10'27 8.72 x 1018
The impact using TEWI, TLTI, GDI 1 and GDI2 are shown for the different system
characteristics, relative to the R22 reference system in Figure 32. Initial general observations are
that the impact of R290 is always lower than that of the R22 system, and for a given system
efficiency, the difference is greater with a high leak. Subsequently, TEWI of the R290 system
remains similar regardless of leak rate, whilst there is considerable variation between TEWI of
the different R22 systems. TLTI closely follows the impact indicated by TEWI, although it is
generally a little lower than TEWI for the R290 systems, and slightly higher for the R22 systems
with high leakage. GDI1 always implies a higher relative impact than TEWI regardless of system
or refrigerant, except in the one case of a low efficiency R22 system with low leakage. Finally,
GDI2 follows the pattern of TLTI where it has a higher relative impact for R22 systems with
large leakage, but an appreciably lower impact for all R290 systems. The variations between the
different measures of global warming impact with differing system characteristics are examined
further by comparing the relative variation of each against that of TEWI. To evaluate the data
from Table 23 in this context, it is presented in Figure 33 as the relative change in impact for
each system, compared to the change o f TEWI (e.g., TLTI !T L T lref ^ TEW I! TEWI ref), where
different trends emerge depending upon the measure.
107
1.4
■ TEWI □ TLTI
§ g d i:
R22 | R290 R22 | R290 R22 | R290 R22 | R290
h ^ i leak [ low leak hgileak b w leak
hjgh efficiency b w efficency
Figure 32: Climate impacts fo r different system characteristics relative to R22 reference system
In terms of TLTI, comparison of results for systems using R290 shows that the relative impact is
almost identical to that o f TEWI regardless of leak rate and efficiency. The R22 systems have
TLTI greater relative to TEWI for the high leak cases and less than TEWI for low leakage,
although the variation is more pronounced for the low efficiency systems. When comparing TLTI
for the C 0 2 contribution only of the different systems (not shown), the relative change against
TEWI was also negligible, as with R290. These observations demonstrate that TLTI consistently
emphasises the rapid global wanning contribution of a refrigerant's high F and r .
Initial observations for GDI1 are that the relative change in impact is nearly always greater than
TEWI regardless of refrigerant, although for a given system efficiency the contribution is less
when leak rates are low. Similarly, the impact relative to TEWI is greater when system efficiency
is low. In general, these findings indicate that GDI1 amplifies the contribution of both refrigerant
and C 0 2 emissions, which is expected given that GDI1 is a direct escalation of AGTP. The one
exception is the low leakage, low efficiency R22 system where the relative GDI 1 is less than
TEWI, which suggests that the contribution of the refrigerant is still given a greater weighting
than CO 2 .
GDI2 produces similar results to TLTI for R290 systems where there is negligible difference
between GDI2 and TEWI, which is consistent with its direct dependence on TLTI. However,
results for the R22 systems reveal somewhat different results since GDI2 demonstrates a lesser
impact than TEWI for three of the four cases. O f these the high efficiency, low leakage system
exhibits a veiy large reduction relative to TEWI, and conversely the low efficiency, high leakage
system is the only case where the relative impact is greater than TEWI. These observations
108
suggest that GDI2 dampens the contribution of C 0 2, but overall there is a more distinct indication
of which system designs are favourable in terms of global warming impact.
0 .1 5 i---------------------
□ TLTI 0GDI1 SG D I2
B 0.10
2
I 0.05
gj 0 .0 0 vm' !
R22I R290 R 23 R290 R22 R290 R 23 R290 i
2 -0.05
-S - 0 .1 0 -
high leak | Icrn' leak high leak | low leak

high efficiency low efficiency
-0 20
Figure 33: Climate impacts fo r different system characteristics relative to R22 and R290 TEWI
To summarise, this Section has presented examples where the proposed alternative measures for
evaluating the climate change impact o f refrigerating systems, using a high F and t refrigerant,
and a low F and r refrigerant have been illustrated. A newly derived measure known as AGTP
was employed to develop various climate impacts because (a) for most refrigerants it avoids
reliance upon accuracy of evaluating C 0 2 , (b) less reliance upon selection of appropriate time
horizons, (c) it is simple to extend to a damage index, and (d) it?s meaning is comprehensible to
non-climatologists. These indices are TLTI, GDI1 and GDI2 as detailed in Section 3. Through
evaluating selected systems, the relative climate impact using these measures was found to be
similar, although they tended to emphasise the effect of refrigerant over the C 0 2 contribution.
Lastly, one crucial feature is that the choice o f how global warming impact is interpreted can
change the rating order of different system. In a practical sense, a TEWI comparison of two
different system designs (including different refrigerant choice) may imply one of the options
produces least global warming impact, whilst use of TLTI, GDI1 or GDI2 may demonstrate the
other option has the smallest impact. This disparity can be used advantageously, in that greater
confidence is gained when an option produces the lowest impact with all four measures.
4.8 Use and advantages of methodology

The comprehensive method presented here can be o f use in two respects. The first main use is for
accurate evaluation of different designs of equipment, whether it is refrigerant or component
109
selection. This approach can be extended to GHG emission optimisation of equipment design,
installation and optimisation, for manufacturers, installers, service companies and system
operators. Secondly, it can be used for technology assessment purposes, where the overall impact
of prescribing certain measures within policies can be evaluated in terms of reduction of CO 2 and
C 0 2 equivalent emissions.
With the increasing attention paid to achieving reduced environmental impact, industry is under
increasing pressure to employ equipment and processes that are low in GHG emissions. For
example, ‘‘high street brand” companies are increasingly specifying high efficiency, low-GWP
cooling equipment, and for servicing companies to have strict refrigerant containment work
procedures in place. The approach detailed in the present study would allow those involved to
analyse designs or procedures and demonstrate a low impact, or to help determine an optimum
design or procedure. Examples of this could include:
- best distribution of funds for system components and materials,
- most suitable production and installation procedures,
- preferred refrigerant selection,
- optimum design of equipment for anticipated temperature and load profiles,
- favourable sendee and maintenance regime,
- most effective route for disposal of refrigerant and equipment.
Using a more transparent approach such as this would help eliminate bias from '"generalised”
results being presented as actuality.
Studies used for policy reasons, such as estimation o f national GHG emissions for RAC sectors
(e.g., March 1999, Haydock et al, 2003) employ a very simplistic approach to estimate impacts.
These generally draw upon studies that are more specific, where systems are evaluated for
different refrigerant selection, and associated generic system designs (e.g., AFEAS, 1997, etc.).
Evaluations using the model show such a wide variation in TEWI results for a single system, that
a few small changes in assumptions can easily reverse the rating of two different refrigerants or
system types. This suggests that using the basic TEWI methodologies as detailed within such
reports is erroneous and are unlikely to lead to valid findings. Consequently, a more analytic
approach that employs statistical frequencies for occurrence of different influencing assumptions
should be use to produce a range of TEWI results for a given sector. If a sector can be broken
down into a range of different designs, practices, etc., a thorough and reliable TEWI evaluation
can be made across the range.
110
SECTION 5: CONCLUSIONS
5.1 Summary of work

The current investigation relates to emissions of GHGs from RAC systems and their effect on
climate change. Review of the literature found that the current calculation procedures as used by
manufacturers, installers and end-users, van in name, scope and detail. Most calculation
procedures are left open to many assumptions, and reported evaluations tend to employ
assumptions that are either not stated, or made to suit some predetermined conclusion.
To address the various shortfalls, a detailed emissions model was constructed, which included
numerous sources of emissions associated with the equipment lifetime; an exhaustive list of these
are listed in Appendix 1 (Table 24). The following aspects examined in terms of their effect on
overall lifetime emissions: size and range of equivalent emissions from different sources,
influence of external conditions, effects of system ageing, lifetime emissions from power stations,
design and selection of equipment and components, refrigerant selection, repair and maintenance
o f system and topping up of refrigerant, behaviour during servicing and refrigerant and material
handling. In addition, the sensitivity of assumptions was examined by considering the range of
anticipated emissions from a particular source.
Lastly, because GWP is known to have shortfalls in terms of accurately describing the
implication of the overall TEWI value, three alternative measures were adopted to interpret the
impact of GHGs emissions. The impact resulting from eight different system designs w ere
compared.
5.2 Key findings

Under certain circumstances, emissions associated with equipment production and servicing
activities can represent a significant proportion of total emissions. Omission of these aspects in
the standardised approach may result in significantly different TEWI. Assumptions about
operating parameters also have a large influence on the output, for example, load profile and
sensitivity’ of performance to charge.
Accurate evaluation of TEWI for a particular system requires detailed knowledge of

environmental conditions and how7 the system behaves under the anticipated range of conditions,
system control strategy, design o f evaporator, condenser, compressor and pipew ork, know ledge
o f service and refrigerant handling processes and behaviour, and information of production of
111
materials, component/assembly manufacturing and disposal procedures. These revealed a number
of important observations to be considered in assessments of RAC equipment, as summarised
below.
- Distribution o f emission size. The majority of the contribution (35 - 95%) normally occurs
from system operating energy and the second most significant (3 - 58%) is normally leakage
from the system. Refrigerant emissions from servicing activities during in-use stage may also
be of similar size. The contribution o f EOL and SOL may vary between <1% to over 10% of
the total. Proportions of the total contribution attributed to different sources vary according
to equipment size, where relative contribution from energy consumption and system leakage
increase with larger equipment (67% for 4 kW to 91% at 58 kW).
- Equipment design. Component selection has an influence on annual leakage rates (2 - 20%),
and as a consequence annual energy consumption (±5%) and application temperature (up to
20 K). An increase of evaporator and condenser sizes can reduce emissions (by 10% when
doubling the size) from energy consumption or increase (by 50%) by halving their size. Using
larger exchangers also raises the system refrigerant charge, in which case greater refrigerant
emissions can also offset the benefits of improved system efficiency. Increasing SOL
emissions through larger exchanger size has a negligible influence on total emissions (< 1 %),
as does recycling equipment materials (1 - 2%). The sensitivity of the system to sub-optimal
charge can affect both emissions from energy production and application temperature; if the
system can achieve design temperatures, emissions increase with higher leakage and
sensitivity to charge level (±2 0 %), whereas high sensitivity may result in greater emissions
for lower leakage rates, but applications temperature moves further from the design value. A
smaller compressor (in terms of sw ept volume) always helps reduce emissions, although the
influence of less materials, oil and refrigerant charge is minor (<1%). How ever, below a
certain size, the compressor becomes too small to match the capacity7 under all conditions.
The effect of refrigerant selection influences emissions in two respects. Contribution from
refrigerant emissions is proportional to the refrigerant GWP. Contribution from energy
consumption is largely dictated by refrigerant properties, but also the basis for comparison is
considered crucial. The ranking between refrigerants can reverse when the basis is changed
(±40%). Similarly, comparison refrigerants on a different basis will result in some
refrigerants not achieving design application temperatures.
112
- External influences. The evaluation of energy consumption over a more precise range of
ambient temperatures provides increasingly accurate results for subsequent emissions
(±100%). Anticipated changes in the UK CO: emissions per kWh of electricity result in a
lower lifetime emissions from energy consumption, up to 5%.
- Equipment handling. The manner in which installation, service and maintenance personnel
cany- out refrigerant handling activities affects total emissions, up to ±30%. Increased
frequency o f maintenance can help to reduce emissions, energy consumption and maintain
application temperature.. Where a refrigerant has a higher GWP, changing compressor oil
and replacing filter/driers can cause an increase in total emissions. More frequent top-ups
may be less beneficial than repair plus topping-up because system leakage is higher, but
repairing increases serv icing refrigerant releases which can offset the benefit of lower leak
rate if refrigerant handling practices are lax. Where a low GWP refrigerant is used, topping-
up is the favoured approach. Effects of ageing (compressor wear, exchanger fouling, increase
in leak rate over time) can affect emissions from energy consumption by up to ± 1 0 %, but
also result in the inability to achieve design application temperatures.
- Interpretation o f consequence o f emissions. Comparison based simply on GHG emissions is

pointless unless they are considered in parallel with the state o f application temperatures. An
important conclusion drawn from evaluating the influence of refrigerant handling activities, is
that a balance has to be sought, between the benefits associated with improved system
performance (e.g., topping-up the charge) and the increase in emissions when servicing. The
higher the refrigerant GWP, the more importance this trade-off becomes. Similarly, by using
a refrigerant with very low GWP, the impact associated with peripheral refrigerant emissions
can be neglected and so servicing activities that improve performance ought to be as frequent
as possible.
If a simplistic approach such as BRA (1996) w'ere employed, a significant over-prediction in the
contribution of energy-related emissions (up to 40%), and under-prediction of high-GWP
refrigerant contributions by a factor of 10 occurred. Finally, the findings reported in Section 4
help identify a number of design and operation features that can assist in minimising GHG
emissions over equipment lifetime.
The present study has produced a number of outputs that are considered to be original
contributions to this subject:
113
- A thorough procedure for TEWI evaluation has been formalised, using a detailed modelling
approach, which can demonstrate the influence o f various equipment design parameters on
total emissions. This was achieved by identifying an exhaustive list of different sources of
emissions within the lifetime of equipment and calculations to evaluate them, and a detailed
breakdown of GHG emissions from different sources during the lifetime of the equipment,
along with uncertainty ranges was presented.
- Part of the emissions model employed correlations produced to approximate masses of
system components and thereby estimating emissions associated with their manufacture,
including a factor to account for the construction of the system. Using a system model for in-
use energy7 consumption calculations, the influence of ageing effects was incorporated by
accounting for degradation of compressor performance, exchanger fouling and the effect of
leakage. A comprehensive empirical leakage model was presented, based on number and
selection of different components within the circuit, and the frequency of repair of the leaks.
For peripheral sources, refrigerant emissions during handling procedures for production,
servicing and decommissioning were characterised. In terms of indirect emissions due to
power consumption, a correlation for estimating variable emissions of CO 2 per kWh of
energy7 consumed according to the year, time of year and part of the day, was established
along with upper and lower ranges.
- Novel alternative means to evaluate the impact of GHG emissions from a refrigeration
sy stem were presented, based on absolute global temperature change due to an emission o f a
particular GHG; TLTI, GDI 1 and GDI2. A comparison o f these different interpretations of
impacts from GHG emissions using several different system designs was offered.
- Omissions and errors associated w ith the BRA (1996) method were identified, and the
resulting uncertainty in using the BRA method presented.
The importance of the present study is reflected by considering the current demands for reducing
environmental burden of equipment through national, European and international legislation, and
environmental policies of businesses. Information provided by this study can assist manufacturers
and users of RAC equipment to improve the quality of their environmental assessments when
considering different options. Further, it helps focus attention towards the aspects of equipment
design, construction and operation that can give most benefit when trying to reduce impact due to
GHG emissions. In particular, there exists an extensive choice of refrigerants whose GWP ranges
from zero to several thousand, and properties that dictate a wide range of efficiencies. The
findings of this study thereby assist with identifying the most suitable refrigerant options.
114
5.3 Limitations and further work
The first consideration is that sub-models associated with the overall emissions model could
always be made more accurate, and observations in the differences in model output when
including “peripheral'’ aspects (such as fouling) confirms this. In particular, improved models
should include:
- More accurate models for emissions from manufacturing and equipment installation.
- Better characterisation of system leakage, including the effects o f external temperatures and
operating conditions on leakage.
- Improved system performance model, particularly with respect to transient effects, also
including effect of evaporator frosting and subsequent defrost strategy.
- Ability to account for differences in associated system designs, such as indirect circuits,
multiple evaporators/compressors, and thermal storage.
- Other than simple on/off cycling, the effect of other control features such as pressure
switches, defrost timing, evaporator/head pressure regulating valves should be accounted for.
- Ancillary system components such as desuperheaters and suction-liquid heat exchangers that
may produce changes in energy’ consumption should be included.
- Inclusion of a more precise “bin” type approach for improved ambient temperature ranges
and calculation of load profiles.
- Extension of the model to include for blended refrigerants, which could be useful for
predicting changes in impacts due to preferential leakage of refrigerant components with
different GWP, and effect on system performance.
- Specifying actual components used within a particular system, and also validating the
performance model using measurements from that, or a similar system.
A major addition to the model would be to integrate a cost model. This would enable cost-
effectiveness analysis of different emission reduction measures. This would help to pinpoint areas
of system design or operation wbere expenditure should focus, and it would enable economic
optimisation to achieve maximum emission reductions.
As a final observation, this study has shown that the use of TEWI for the usual purpose of
comparing refrigerants has major uncertainties associated with it, and results can be highly
sensitive to assumptions. Therefore, the most reliable approach for conducting a TEWI study
would be to make it retrospective. Following the study of Benstead (1996), two different types of
115
systems should be installed and operated over a given lifetime, although a statistically significant
number of each system types would have to be used. Production and installation would be closely
monitored and evaluated, as would the EOL stage activities. Moreover, the performance (and
corresponding local conditions) of each system within the population should be recorded as well
as all service and maintenance activities. Full evaluation of individual systems within each
population would provide a genuinely meaningful comparison, and additionally, could be used
also to help validate models such as this.
116
REFERENCES
AIA. Condenser and evaporator product catalogues. Asarum Industri AB, S Industrivagen 2-4,
SE-374 50, Asarum, Sweden. Undated.
Aprea, C., Greco, A. An experimental evaluation o f the greenhouse effect in R22 substitution.
Energy Conserv. Mgmt. Vol. 9, pp. 877-887. 1998.
Aprea C., Moncada, L. 0 ., Giudice G. R. A simplified criterion for evaluating the minimum
contribution to the ‘‘greenhouse effect" of a vapour compression system. Proc. IIR, 20th Int.
Congr. Refrig. Sydney, Australia. 1999.
Aprea, C., Mastrullo, R. An experimental evaluation of the vapour compression plant

performances in presence of R407C leaks using an electronic expansion valve. Applied Thermal
Engineering, Vol. 22, pp. 161-171. 2002.
ARAP. Appendix A: Embodied energy7 and GWP o f fugitive emissions of fluorocarbons. Alliance
for responsible atmospheric policy, http://www.arap.org/adlittle/appendixa.html. Last accessed
March 2004.
ASHRAE. Refrigeration Handbook. American Society for Heating, Refrigeration and Air
conditioning Engineers. Atlanta, USA. 2002.
AutoLPG. Table 4. www.autolpg.com/autocom.htm Last accessed 22nd December 1998.
Ayers, J. Personal communication. Parascnse study on refrigerant leaks. Parasense

Environmental Products Ltd, Gloucester, UK. November 2000.
Beggs, C. B. A method for estimating the time-of-day carbon dioxide emissions per kWh of
delivered electrical energy in England and Wales. Building Services Engineering Research and
Technology7, Vol. 17, No. 3, pp. 127 - 134. 1996.
Banks, R. E., Sharrat, P. N. Environmental impacts of the manufacture of HFC-134a.

Department of Chemical Engineering, UMIST, Manchester. 1996.
Banks, R. E., Clarke, E. K., Johnson, E. P., Sharratt, P. N. Environmental aspects o f fluorinated
materials: Part 3. Trans. I. Chem. E., Vol. 76, Part B. 1998.
Bamabe, F., Khalifa, K., Malek, A. Life cycle analysis: first stage in the development of new
eco-tools for air conditioning systems designers. Proc. IIR, 21st Int. Congr. Refrig. Washington,
USA. 2003.
Benstead, R. Audit of TEWI analysis of refrigeration designs for Henley-on-Thames store. EA

Technology7 Services Report, V1684. Capenhurst, UK. 1996.
117
Bitzer. Compressor catalogues, www.bitzer.com. Last accessed September 2004.
Blomhardt. R. Personnal communication. Bitzer Kuhlmaschinenbau GmbH.. Germany. 2002.
Bodyul, O., Dubro, I. V. An ecological approach to a problem o f risk evaluation of halocarbons.

Proc. HR, 21st Int. Congr. Refrig. Washington, USA. 2003.
BRA. Guideline methods of calculating TEWI. Issue 1. British Refrigeration Association, Bourne
End, UK. 1996.
Broadhurst, P. Personal Communication. Linde Refrigeration Ltd, Oxford, UK. January, 2002.
BS 5925: 1991 Code of Practice for Ventilation Principles and designing for natural ventilation.
British Standards Institution, London, UK. 1991.
Butler, D. J. G. Refrigerant leakage and leak detection. Report Ref: Pd255/93, Building Research
Establishment, Watford, UK. 1993.
Buttervvorth, D. An approach to fouling allowances in the design of heat exchangers. Applied

Thermal Engineering. Vol. 22: 755-762. 2002.
Campbell, N. J., McCulloch, A. Climate change implications of manufacturing refrigerants: A

calculation of production energy contents of some common refrigerants. Process Safety and
Environmental Protection: Transactions of the Institution of Chemical Engineers, Part B, Vol. 76,
No. 3, pp. 239-244. 1998.
Cintar, C., Hadfield, M. An environmental evaluation of mechanical systems using

environmentally acceptable refrigerants. Int. J. LCA, Vol. 5, No. 4, pp. 209-220. 2000.
Clodic, D. Zero Leaks: limiting emissions o f refrigerants. ASHRAE, Atlanta, USA. 1996.
Colboume, D. Practical application of hydrocarbon refrigerants: production costs and operational

performance. Proc. IIR, 20th Int. Congr. Refrig. Sydney, Australia. 1999.
Colboume. D., Suen K. O. Expanding the domain of TEWI calculations. Proc. IIR, 20th Int.
Congr. Refrig. Sydney, Australia. 1999.
Colboume. D., Suen K. O. Assessment of performance of hydrocarbon refrigerants. Proc. IIR

Gustav Lorentzen Conf., Purdue, USA. 2000.
Colboume. D., Suen K. O. Application of Global Temperature Potential to Refrigerating

Systems: Total Lifetime Temperature Impact (TLTI). Proc. IIR Gustav Lorentzen Conf.,
Glasgow, UK. 2004.
118
Davies, T. W., Caretta, O. A low carbon, low TEWI refrigeration system design. Applied
Thermal Engineering, Vol. 24, pp. 1119 - 1128. 2004.
Danfoss. Compressor and valve catalogues, www.danfoss.com. Last accessed September 2004.
Douglas, J.D., Braun, J. E., Groll, E. A., Tree, D. R. A cost-based method for evaluating
alternative refrigerants applied to R22 systems. Int. J. Refrig., Vol. 22, pp. 107-125. 1999.
DTI. Digest of United Kingdom Energy Statistics. Government Statistical Service, Department
for Trade and Industry , London. 1998.
DTI. Energy Paper No. 68 (EP68). Department for Trade and Industry , London. 2000.
EN 378: 2000 Part 1. Refrigerating systems and heat pumps - safety' and environmental
requirements. BSI, London, 2000.
ETSU. Good practice guide 178: Cutting the cost of refrigerant leakage. Energy and Technology
Support Unit, DETR, Harwell, UK. 1997.
Farzad, M., O'Neal, D. L. Influence of the expansion device on air-conditioner system

performance characteristics under a range of charging conditions. ASHRAE Transactions:
Research, Paper No. 3622, pp. 3-13. 1992.
Fischer, S. K., P. J. Hughes, P. D. Fairchild, C. L. Kusik, J. T. Dieckmann, E. M. McMahon,

and N. Hobday. Energy and global warming impacts of CFC alternative technologies. Alternative
Fluorocarbons Environmental Acceptability Study and U.S. Department of Energy7, Washington,
USA. 1991.
Fischer, S. K., Tomlinson, J. J., Hughes, P. J. Energy' and global warming impacts of not-in-kind
and next generation CFC and HCFC alternatives. Alternative Fluorocarbons Environmental
Acceptability Study and U.S. Department of Energy', Washington, USA. 1994.
Frankhauser, S. The social costs of greenhouse gas emissions: an expected value approach.
Energy J., Vol. 15, pp. 157-184. 1994.
Frischknecht, R. Environmental relevance of natural refrigerants, life cycle assessments of heat

pumps and refrigerating systems. ESU Services. Report for the Federal Office of Energy,
Switzerland. December 1999.
Fuglestvedt, J. S., Bemtsen, T. K., Godal, O., Sausen, R., Shine, K. P., Skodvin, T. Metrics of
climate change: assessing radiative forcing and emission indices. Climatic Change, Vol. 58, pp.
267-331.2003.
119
Galloway, J. E., Goldschmidt, V. W. Relationship between refrigerant charge, superheat,
capacity and efficiency of a specific heat pump. Proc. Int. Congr. Refrig., Quebec, Canada.
1991.
Garland, N. P. Sustainable Design of Hydrocarbon Refrigerants Applied to the Hermetic

Compressor. PhD. Thesis. School of Design Engineering and Computing. Bournemouth
University, Bournemouth. 2004.
Godal, O. The IPCCs assessment of multidisciplinary issues: the case of greenhouse gas indices.
Climatic Change, Vol. 58, pp. 243-249. 2003.
Gover, M. P., Collings, S. A., Hitchcock, G. S., Moon, D. P., Wilkins, G. T. Alternative road
transport fuels - a preliminary life-cvcle study for the UK. Volume 2. ETSU, Harwell, UK.
1996.
Gopalnaravanan, S., Rolotti, G. D. Total Equivalent Warming Impact of R-22 Alternatives in

Air-Conditioning and Heat Pump Applications. ASHRAE Transactions, Vol. 105, No. 1, pp.
1228-1236. 1999.
Hammond, A. L., Rodenburg, E., Moowmaw, W. Commentary on global warming indices.

Nature, Vol. 347, pp. 705-706. 1990.
Hansen, P. E., Gustafsson, V. Total environmental impact of a small hermetic compressor. Proc.
Int. Refrig. Conf., Purdue, USA. 1994.
Haydock, H., Adams, M., Bates, J., Passant, N., Pve, S., Salway, G., Smith, A. Emissions and
projections of HFCs, PFCs and SF6 for the UK and constituent countries. Final Report Prepared
for the Department Of Environment, Food And Rural Affairs. AEA Technology, Abingdon, UK.
July, 2003.
Heikkila, K., Fahlen, P. Evaluation of environmental impacts of air-conditioning systems at the

design stage. Proc. IIR Int. Congr. Refrig. . Washington, USA. 2003.
Hwang, Y., Jin, D-H., Radermacher, R. Comparison of Hydrocarbon R-290 and Two HFC
Blends R-404A and R-410A for Medium Temperature Refrigeration Applications. Final Interim
Report. CEEE, Department of Mechanical Engineering, University' of Mary land. February 2004.
Incroprea, F., De Witt, D. Fundamentals of Heat and Mass Transfer. 3rd Ed., Wiley, Singapore.
1990.
1PCC (Intergovernmental Panel on Climate Change). First Assessment Report. (Includes:

Scientific Assessment of Climate change - Report of Working Group I, Cambridge University'
Press, UK; Impacts Assessment of Climate Change - Report of Working Group II, Australian
120
Government Publishing Service, Australia; The IPCC Response Strategies - Report of Working
Group 111, Island Press, USA.) 1990.
IPCC (Intergovernmental Panel on Climate Change). Second Assessment Report. (Includes:

Climate Change 1995: The Science of Climate Change, Cambridge University Press, UK;
Climate Change 1995: Impacts, Adaptations and Mitigation of Climate Change: Scientific-
Technical Analyses, Cambridge University7Press, UK; Climate Change 1995: Economic and
Social Dimensions of Climate Change, Cambridge University Press, UK.) 1995.
IPCC (Intergovernmental Panel on Climate Change). Third Assessment Report. (Includes:

Climate Change 2001: The Scientific Basis, Cambridge University Press, UK; Climate Change
2001: Impacts, Adaptation & Vulnerability', Cambridge University Press, UK; Climate Change
2001: Mitigation, Cambridge University7Press, UK.) 2001.
IPCC (Intergovernmental Panel on Climate Change)/TEAP (Technology and Economic

Assessment Panel). Special Report on Safeguarding the Ozone Layer and the Global Climate
System: Issues related to Hvdrofluorocarbons and Perfluorocarbons. Cambridge University
Press, Cambridge, UK. 2005.
ISO 14040:1997 Environmental management - Life cycle assessment - Principles and

framework. International Standardisation Organisation. Geneva, Switzerland. 1997
James, R. Personal Communication. Rechem Ltd, UK. February7 1999.
Johnson, E. Global warming from HFC. Env. Impact Asses. Rev., Vol. 18, pp. 485 - 492. 1998.
Jurgensen, H. Personnal communication. Danfoss Compressors, Denmark. 2004.
Kruse, H. Is the TEWI number suitable for evaluating the combined global warming effect of
refrigeration and heat pump systems? HVAC&R Research J., Vol. 4, No. 3. 1999.
Lashof, D. A., Ahuja, D. R. Relative contributions o f greenhouse gas emissions to global

warming. Nature, Vol. 344, pp. 529-531. 1990.
Lundqvist, P. Personal communication on compressor performance tests. Department of Energy

Technology (KTH), The Royal Institute of Technology, Stockholm. 2004.
Manne, A. S., Richels, R. G. An alternative approach to establishing trade-offs among

greenhouse gases. Nature, Vol. 5, pp. 675-677. 2001.
March. UK Emissions of HFCs, PFCs and SF6 and potential emission reduction
opportunities.For Global Atmosphere Division, DETR, London. 1999.
121
McCulloch, A., Campbell, N. J. The climate change implications of producing refrigerants. Proc.
HR Gustav Lorentzen Conf., Oslo, Norway. 1998.
McCulloch, A.; Lindley, A. A. From mine to refrigeration: A life cycle inventory analysis of the
production of HFC-134a. Int. J. Refrig., Vol. 26, No. 8. 2003.
McCulloch, A.; Midgley, P. M., Ashford, P. Releases of refrigerant gases (CFC-12, HCFC-22
and HFC-134a) to the atmosphere. Atmospheric Environment, Vol. 37, pp. 889-902. 2003.
Norgate, T. E., Rankin, W. J. Life cycle assessment of copper and nickel. Proc. Minprex, Int.
Conf. Materials Processing, pp. 133-138. Australia. September 2000.
NRS. Refrigeration component catalogues, www.nrs.co.uk. Last accessed September 2004.
O'Neill, B. C. The jury is still out on global warming potential. Climatic Change, Vol. 44, pp
427-443. 2000.
Papasawa, S. Moomaw, W. Life-cvcle global warming impact of CFCs and CFC substitutes for
refrigeration. J. Industrial Ecology, Vol. 1, No. 4. 1998.
Philipp, J., Kraus, W. E., Quack, H. Numerical simulation of refrigeration cycles with capillary’
tubes in on/off cycling mode. Proc. IIR Gustav Lorentzen Conf., Purdue, USA. 2000.
Pira. Ecopackager. Streamlined Life Cycle Assessment Model. Pira Ltd, UK. 2001.
Radford, P. The benefits of refrigerant inventory control. Proc. Conf. Inst. Refrig., London, UK.
1998.
Radford, P. Detection, causes and nature of refrigerant leaks. Proc. Inst. Refrig., London, UK.
1993.
Reilly J., Prinn R., Hamisch J., Fitzmaurice, J., Jacoby, H., Kliklighter, D., Melillo, J., Stone, P.,
Sokolov, A., Wang. Multi-gas assessment of the Kyoto Protocol. Nature, Vol. 401, October,
1999.
Reilly, J. M., Richards, K. R. Climate change damage and the trace gas index issue. Environ,
Resour. Econ., Vol. 3, pp. 41-61. 1993.
Ritter, T. J. Personal communication. Calor Gas Ltd, Warwick, UK. 2004.
SAEFL. Life cycle inventories for packaging. Environmental Series No. 250/1 Waste. Swiss
Agency for the Environment, Forests and Landscape, Berne, Switzerland. 1998.
122
Sand, J. R., Fischer, S. K., Baxter, V. Energy and global warming impacts of HFC refrigerants
and emerging technologies. Oak Ridge National Laboratory. For Alternative Fluorocarbons
Environmental Acceptability Study (AFEAS) and Department of Energy. USA. 1997.
Sand, J. R., Fischer, S. K., Baxter, V. Comparison of TEWI of alternative fluorocarbon

refrigerants and technologies in residential heat pumps and air conditioners. Oak Ridge National
Laboratory. For Alternative Fluorocarbons Environmental Acceptability Study (AFEAS) and
Department of Energy. USA. 2002.
Searle. Evaporator and condenser catalogues, www.searle.co.uk. Last accessed September 2004.
Shine, K. P., Fuglestvedt, J, S., Hailemariam, K., Stuber, N. Alternatives to the global warming
potential for climate impacts of emissions of greenhouse gases. Climatic Change, Vol. 65, No. 4.
2004.
Smith, S. J. The evaluation of greenhouse gas indices. Climatic Change, Vol. 58, pp. 261-265.
2003.
Somerscales, E. F. C. Fundamentals of corrosion fouling. Experimental Thermal and Fluid

Science. Vol. 14. pp. 335-355. 1997.
UNEP (United Nations Environment Programme). Report of the Task Force on Destruction
Technologies. Technology and Economic Assessment Panel, Volume 3B. Nairobi, Kenya. April
2002 .
UNEP. Report of the refrigeration, air conditioning and heat pumps committee. United Nations
Environment Programme, Ozone Secretariat, Nairobi, Kenya, 1999.
Ure, Z. Effective control, energy efficiency and system diversification influence on TEWI. Proc.
Int. Congr. Refrig., The Hague, Netherlands. 1995.
Wigley, T. M. L. The Kyoto Protocol: C02, CH4 and climate implications. Geophys. Res. Lett.,
Vol. 25, pp. 2585-2288. 1998.
Zhang, M., Reistad, G. M. Analysis of energy conservation systems, including material and
global warming aspects. HVAC & R Research, Vol. 4, No. 1. ASHRAE, USA. 1998.
Zhelezny V. P., Zhidkov, V. V., Butner, A. G., Muravia, E. S. Ecological and energetical aspects
of changing over refrigerant equipment in Ukraine to alternative refrigerants. Proc. IIR Natural
Working Fluids Confi, Aarhus, Denmark. 1996.
123
APPENDICES
Appendix 1: Compilation of sources found to affect emissions
Table 24: Exhaustive list o f sources o f emissions and influencing parameters

Stag Activity Source (and type) Influencing parameters
e
Material processing facilities (CO2 ) Material type, component types and mass,
recycling
Material
Refrigerant process plant (CO:, R) Refrigerant type, lifetime used mass,
production
recvcling/EOL incineration
Refrigerant distribution (R) Lifetime used mass, means of distribution
Components Manufacture, assembly (CO>) Factory “efficiency”
SOL
Purging/leak testing (R) Type of gas. Hose/manifold size
Installation Charging (R) Hose/manifold size
Construction/assembly (CO2 ) Equipment size, no. connections

Equipment movement (CO2) Mass of equipment
Transport
Site visits (CO2 ) Size of installation
Compressor energy (CO2) Power station emission factor, component and
refrigerant selection/design, (efficiency,
pressure losses, heat transfer, application
thermal mass, application temperature, ambient
temperatures, compressor wear, exchanger
Operation fouling, system charge sensitivity,
servicing/maintenance frequency
Ancillary device energy (CO2 ) Power station emission factor, component
selection, system cycling times
In-
use Refrigerant leakage (R) Component numbers, type, repair frequency
System venting (R) Behaviour
Refrigerant recovery (R) Use of recovery' machine, hoses/manifold
Servicing/
System evacuation (R) Oil charge, solubility
maintenance
Charging (R) Hose/manifold size
Purging/flushing/leak test (R) Type of gas, Hose/manifold size
Transport Site visits (CO2 ) Distance, visit frequency
System venting (R) Behaviour
Decom
Refrigerant recovery (R) Use of recovery machine, hoses/manifold
missioning
Dismantling (CO2, R) Oil charge, solubility
EOL Transport Equipment movement (CO2 ) Mass of equipment
Plant emissions (CO2, R) Mass of refrigerant, mass of oil, incineration
Incineration
efficiency
Disposal Material/refrigerant production Recycling
124
PART 2: DEVELOPING A NEW SYSTEM BALANCING TECHNIQUE FO R
M IXTURE REFRIGERANTS
125
NOMENCLATURE
Letters
a a constant (-)
A area (m2)
A
TEV area of TEV orifice (m2)
A total design surface area of heat exchanger (m2)
tot.des
b a constant
B basic rating (kW K'1)
^ cond condenser basic rating (kW K'1)
b dsh basic rating for desuperheating region (kW K 1)

D evaporator basic rating (kW K'1)
evap
Bo boiling number (-)
K basic rating for subcooling region (kW K"1)
bsh basic rating for superheat region (kW K 1)
c a constant
C capacity rate (kW K'1)
cd discharge coefficient (-)
c HTF HTF capacity rate (kW K 1)
^ max maximum capacity rate (kW K'1)
C_-_
rmn minimum capacity’ rate (kW K"1)
COP coefficient of performance (-)

Cp specific heat at constant pressure (kJ kg'1 K '1)
CPhtf specific heat of HTF (kJ kg'1 K'1)
CPr refrigerant apparent specific heat (kJ kg'1 K 1)
C 'r refrigerant apparent capacity’ rate (kW K'1)
C TEl' TEV constant (-)
Cv specific heat at constant volume (kJ kg"1 K'1)

D tube diameter (m)
Do diameter of outside tube (m)

F* function to account for mixture effects in condensation (-)
F2 function to account for mixture effects in condensation (-)
FJR parameter for mixture heat transfer coefficient (-)
Fme mixture correction factor for heat transfer coefficient (-)
Fr Froude number (-)
F* two-phase enhancement factor (-)
Fun mixture correction factor for heat transfer coefficient (-)
Fix}, condensing heat transfer vapour quality factor (-)
F { x}2 condensing heat transfer vapour quality factor (-)
g gravitational constant (m s'2)
G mass flux (kg s'1m'2)
Gv mass flux of vapour (kg s'1m'2)
h heat transfer coefficient (W m'2 K '1)

H enthalpy (kJ kg'1 K"1)
H disch enthalpy at compressor discharge (kJ kg'1 K '1)
hHTF HTF heat transfer coefficient (W m'2 K'1)
V, heat transfer coefficient between the liquid interface and the liquid (W m'2 K'1)
K v heat transfer coefficient between the liquid interface and the vapour (W m'2 K'1)
h* pool-boiling heat transfer coefficient (W m"2 K'1)
K refrigerant heat transfer coefficient (W m'2 K'1)
enthalpy at compressor suction (kJ kg'1 K'1)
k thermal conductivity (W m'1 K'1)

m mass flow rate (kg s'1)
m HTF HTF mass flow rate (kg s'1)
™r refrigerant mass flow rate (kg s'1)
127
n an index (-)
N number (-)
N crt number of tubes or circuits (-)
nG temperature glide index (-)
N h number o f elements based on enthalpy (-)
N ju number of transfer units (-)
Nu Nusselt number (-)
** number of elements based on vapour quality (-)
P pressure (Pa)
Ap pressure difference/drop (Pa)
Po saturation pressure at 0°C (Pa)
P atm atmospheric pressure (Pa)
P cr,mvc critical pressure of more volatile component (Pa)
Pime ideal mean effective pressure (Pa)
Pmf pressure of mimicking fluid (Pa)
Pr Prandtl number (-)
128
q heat flux (kW m"2)
o rate of heat flow, capacity' (kW)
Q c. comp compressor condensing capacity (kW)
Q c.co n d condenser capacity’ (kW)
O'c.cond
~ condenser capacity at standard rating conditions (kW)
Q c ,c 1u condensing unit condensing capacity (kW)
Q c ,e ! u evaporating unit condensing capacity (kW)
Q d sh capacity of desuperheating region (kW)
Q e.com p compressor evaporating capacity (kW)
Q e ,c i u condensing unit evaporating capacity (kW)
Q e e iu evaporating unit evaporating capacity (kW)
Q e,e\’ap evaporator capacity' (kW)
O e.e\'ap
’ evaporator capacity at standard rating conditions (kW)
Q e jm - TEV evaporating capacity' (kW)
Q sc capacity- of subcooling region (kW)
Q sh capacity of superheating region (kW)
Q sp j capacity of single-phase liquid region (kW)
Q s p .V capacity of single-phase vapour region (kW)
Q m total heat exchanger capacity (kW)
Q *
capacity of two-phase region (kW)
capacity rate ratio (-)
R-cyl clearance volume ratio (-)
R surf ratio of outside (HTF) to inside (refrigerant) surface area (-)
129
T temperature (°C)
T>P bubble-point temperature (°C)
Tc condensing temperature (°C)
^ c (in ) condenser approach temperature difference (K)
Tc,mp condensing mid-point temperature (°C)
Tcold temperature o f cold fluid stream (°C)
Tdisch discharge temperature (°C)
T*P dew-point temperature (°C)
condenser desuperheat (K)
T. evaporating temperature (°C)
^ e iin ) evaporator approach temperature difference (K)
T.* p evaporating mid-point temperature (°C)
ATg temperature glide (K)
adjusted temperature glide (K)
1Thot temperature of hot fluid stream (°C)
Thtf HTF temperature (°C)
AT(m) approach temperature difference (K)
ATlw) approach temperature difference at standard reference conditions (K)
ATmm(1„) minimum approach temperature difference (K)

<
**
saturated temperature difference due to pressure loss (K)
Tr refrigerant temperature (°C)
L temperature at saturation line (°C)
ATsc condenser subcooling (K)

<1
evaporator superheat (K)
Tsh(out) evaporator outlet superheat temperature (°C)
suction superheat (K)
1TEVfin) refrigerant temperature into TEV (°C)
T
1 wall temperature of tube wall surface (°C)
n velocity (m s'1)
U overall heat transfer coefficient (W m'2 K '1)
Vsw compressor swept volume flow rate (m3 s'1)
Wcomp compressor power (kW)
Wfr frictional power (kW)
x vapour quality (-)
x e(m) evaporator inlet vapour quality' (-)
X tt Lockhart-Martinelli parameter (-)
Y, mole fraction of the more volatile component in liquid (-)
Yv mole fraction of the more volatile component in vapour (-)
z distance, length (m)

z crt length of single refrigerant circuit (m)
Greek
P contact angle (°)
s effectiveness (-)
K two-phase multiplier (-)
r ratio of specific heats (-)
Tirol volumetric efficiency (-)
A friction factor (-)

A, liquid-only friction factor (-)
A. vapour-only friction factor (-)
P viscosity' (Pa s)
M molar mass (kmol k g 1)

r temperature parameter (K 1)
3 void-fraction parameter defined in text (-)
P density (kg m'3)
<j surface tension (N m'1)

V void fraction (-)
s correction factor (-)
131
Subscripts
1 condition at evaporator outlet/compressor inlet,or numerical reference
2 condition at compressor outlet/condenserinlet, or numerical reference
3 condition at condenser outlet/TEV inlet, or numericalreference
4 condition at TEV outlet/evaporator inlet,or numerical reference
bp bubble-point
c condensing
cond condenser
comp compressor
c !u condensing unit
dp dew-point condition
dsh desuperheat
e evaporating
evap evaporator
e !u evaporating unit
fr frictional
g gravitational
HTF heat transfer fluid

(in) inlet condition
I liquid
M momentum
max maximum
min minimum
(out) outlet condition
s saturation
sc subcooling
sh superheat
suet suction
TEV TEV
tp two-phase
v vapour
Y mole fraction of the more volatile component
SECTION 1: INTRODUCTION
1.1 New refrigerants

With the phase-out of CFCs and HCFCs a large number of substances have been proposed as
alternatives, including HFCs, PFCs and HCs (UNEP, 2002). Very few of these new substances
possess thermodynamic properties that match the prohibited fluids. Therefore, refrigerant
producers have mixed these substances to obtain likened properties, which are intended to provide
similar system operating characteristics.10
There are no particular technical reasons for developing refrigerant blends, except for ease of
application and use in existing systems. Historically, refrigerant mixtures have been used in niche
applications and experimental investigations for many years (e.g., Hasleden and Klimek, 1957)
but it is only since the beginning of the phase-out of CFCs that zeotropic blends have become
commonplace commercially. It is seen that over three-quarters of commercially available
refrigerants (UNEP, 2002 - Refrigerants Annex) are non-azeotropic refrigerant mixtures. This
fact highlights the importance of studies in this area.
Zeotropic mixtures differ from pure and azeotropic refrigerants in that they exhibit a
“temperature glide" or temperature variation when changing phase. This temperature glide is
caused by the components of the mixture possessing different vapour pressures. When the
mixture begins to change phase, there is preferential boiling (or condensation) of one of the
components with the lower (or higher) vapour pressure. The remaining mixture increases (or
reduces) in temperature until phase-change is complete, and this temperature glide typically
ranges from 0 K to 12 K at normal operation temperatures for most commercially available
zeotropes. The effect of this temperature glide has added a new consideration to the design and
operation of refrigeration components and systems.
1.2 Overview of refrigeration component-system chain

The impact of refrigerant temperature glide on refrigeration components and combination of
components can be observed from a system’s conceptual beginning to its physical operation,
termed the component-system chain. Figure 34 provides an overview of the stages under
10 A mixture, or blend, of two or more refrigerants can be categorised into those that exhibit a
temperature glide (zeotrope), and those that do not (azeotrope). A zeotrope is defined as: a blend
composed of two or more refrigerants whose equilibrium vapour and liquid phase compositions are not
the same. An azeotrope composes of two or more refrigerants whose equilibrium vapour and liquid
phase compositions are the same at a given pressure (ISO 817).
consideration. There are two distinct phases within this chain, which are generally the
responsibility of the product manufacturer, and the utilisation engineers respectively. It is
necessary7to ensure a proper flow of product information throughout the chain, and that a robust
mechanism for this transfer of information is in place. Without this flow of information, the final
operation of a system is unlikely to meet the specifications of the initial design requirements.
Purpose [what the component must achieve]
Design [realising the purpose of the component]
Construction [building the design in real form]

p •=
Performance testing [obtain rating for set of conditions]
Catalogue [use rating data to generate performance data]
Selection [choose components to satisfy- performance requirements]
Balancing [see how selected component behaves with others] " O C /1
fa C/5
CL
Installation [produce a construction with selected components]
Commissioning [confirm operation of sy stem according to balance]
Figure 34: Stages in the component-system chain
The various stages of the component-system chain are described in Figure 34:
- Purpose. A product manufacturer identifies a need for a product. No specific technical
considerations with respect to temperature glide are needed here.
- Design. The design stage is where the purpose of the component is assessed and expressed as
a visualisation of how the purpose will be achieved. Since the design of a component will be
based on specified operating conditions (such as temperature), it is evident that the
component design stage relates to refrigerant temperature glide.
- Construction. This is simply a fabrication of the design where no consideration to
temperature glide is necessary.
134
- Performance testing/rating. The main purpose of testing is for rating. For components in
common use published industry standards usually exist that describe methods for
performance testing. Operating temperatures are implicit in most component performance
assessments and therefore the impacts of temperature glide should also be noted here.
- Catalogue data. Test results and standardised models are applied to generate information
that can be used to select or assess a components ability to perform in a particular
environment. The information normally comes in the form of catalogue data. Interpretation of
the catalogue data is necessary and therefore the rating conditions must work from a
particular basis, such as a reference temperature, hence the implication of temperature glide
at this stage.
- Selection. An utilisation engineer applies the component to a design and eventually the
installation of a system. Selection of a component, chosen from catalogue data to suit the
system operating conditions, is normally based on a number of parameters, including
refrigerant temperature, implying that temperature glide is a consideration here.
- Balancing. Following a selection of various components that form a system an evaluation
technique is required to determine the system characteristics. From the balancing process it is
determined whether the choice of components was suitable to achieve the desired overall
system operating conditions, meeting certain design specifications. Both the balancing and
selection stages are integrated as it may form an iterative process of re-selection. Both stages
are therefore dependent upon consistent design parameters, which include reference
temperatures, implicating the temperature glide.
- Installation. This stage is the physical linking o f the various components with the system
according to the design specification and therefore there is no implication for temperature
glide.
- Commissioning. Once a system has been installed and charged with refrigerant, it will be set
into operating mode and controls, etc., adjusted accordingly. In order to achieve this,
measurements must be made which include temperature measurement and by implication,
temperature glide. However, since there is no numerical procedure, this stage can be
neglected.
The main focus of this study is the transfer of data throughout the component-system chain,
particularly in performance testing, catalogue data, selection and balancing stages, and the
characteristics of zeotropic refrigerants that influence these.
135
Information on conventional methods for performance testing, catalogue rating, component
selection and balancing technique, normally found in industry codes and standards, should be well
understood. Since the key issue is the effect that zeotropic refrigerants and the associated
temperature glide have on these methodologies, a general understanding of the behaviour o f these
refrigerants is also necessary.
1.3.1 Performance testing, rating, selection and balancing

Research literature on component performance, testing, rating and selection is rare since
traditional practices and standards are used, and until recently it has not been necessary7to change
them. Databases from standardisation organisations (such as BSI) were searched for standards
relating to expansion devices, compressors, condensers and evaporators. Selection of components
generally7makes use of catalogue data that has been developed according to the rating procedures.
A review of selected manufacturers literature provided information on the parameters required to
select specific components. For example, nominal air cooled evaporator capacities are based on
entering air temperature and the refrigerant saturation temperature (usually referred as the
approach temperature difference, A7jm)) and correction factors are given for converting standard
dry capacities into nominal capacities for “real'’ or wet conditions.11
Compressor
ISO 917 details a method for conducting measurements on single stage positive displacement
refrigerant compressors, although it could be used similarly for most other types of refrigeration
compressors. Several different test methods are offered by the standard, of w hich two must be
used simultaneously to determine capacity, and the results averaged. ASHRAE 23 specified
similar test methods for reciprocating compressors. The method for presenting compressor rating
data is from ISO 9309. Data are presented for conditions according to standard superheat values.
For each compressor, the data should be fitted to a polynomial equation, to interpolate for fixed
increments in tables.
A selection of compressor catalogues (Prestcold, undated) provided data in both tabular and
graphical formats. For each model, it's capacity, motor power and motor current was provided
11 The reasons for testing units under dry conditions and then applying the correction factors is so
that steady-state conditions can be achieved for testing purposes. Whilst water vapour is
condensing or subliming on exchanger surfaces a transient process is occurring that does not
allow for a constant measurement.
136
according to evaporating and condensing temperature. There was a separate set of data provided
for each refrigerant, and based on specific subcooling and suction temperatures. Correction
factors and charts were offered for adjusting compressor capacity7rating in the case o f non
standard subcooling and superheat settings. Selection is by matching the estimated evaporating
and condensing temperatures of the system with the cooling capacity.
Expansion device
Despite the variety of expansion devices available, only one standard was found (ASHRAE 17),
for testing and rating of thermostatic expansion valves (TEVs) w hich require two test
arrangements. The first is to determine flow7rates for static superheat conditions, wiiere inlet and
outlet pressures are applied to a fixed orifice size corresponding to various phial temperatures.
The second test is for the measurement of mass flow against the degree of orifice opening.
Capacity7rating is obtained from mass flow and refrigerant enthalpies based on certain rating
temperatures. TEV catalogue data (Danfoss, undated) show'ed that valves are selected based on
tabulated capacities defined by nominal evaporating, condensing and liquid entering temperatures
for various pressure drops across the valve orifice.
Condenser
There are standards for performance testing of a variety of different condensers, for example, EN
1117 describes test procedures for liquid-cooled condensers. It specifies a fixed set o f test
conditions under which temperature and flow7rate measurements are made for both refrigerant
and the heat transfer fluid (HTF). Using this data, a standard capacity is determined from which
values for other non-standard conditions could be obtained.
Catalogue data for various refrigerant-air finned-tube condensers was review'ed (Searle, undated).
Condenser selection was based on tabulated capacities for dry conditions, which are dependent
upon a number of variables including the approach temperature difference. The approach
temperature difference used for selection purposes is the difference between the condensing
temperature and ambient temperature. The actual capacity for non-standard conditions was
obtained by applying correction factors to the tabulated capacity'.
Evaporator
Evaporator standards are similar in principle to condenser standards. EN 1118 describes
performance testing and rating methods for various types of liquid coolers. Again, fixed test
conditions are specified for measurements leading to calculation of the standard capacity7. Since
137
the superheat is difficult to maintain at a prescribed value, tests are made across a number of
superheat values (whilst maintaining other specified conditions) and the data interpolated to
determine evaporator capacity at the specified superheat. Non-standard capacities were found
from a general rating equation. Additionally, ASHRAE-25 and EN 328 provide test methods for
air-cooling evaporators that adopt the same basis as EN 1118.
Evaporator selection data (Searle, undated) for horizontal refrigerant-air finned-tube type were
reviewed. The tabulated rated capacity was based on a single fixed evaporating temperature,
adjusted for humid operating conditions, and the actual capacity was found using correction
factors. Correction factors were supplied, to account for evaporating temperature, refrigerant
type and temperature difference (also applicable to wet coils).
Further rating methods

Whilst evaporator and condenser standards provide a basic methodology for estimating heat
exchanger capacity7to conditions other than the test conditions, it only applied to variations of
AT{in). Hence, effects of different construction and operational parameters such as fin spacing,
air flow rate, refrigerant type and air humidity’ on capacities were not accounted for. Since
manufacturers generally offer a wide range of components, performance testing for all possible
configurations would be a costly exercise. Manufacturers therefore tend to produce their own
semi-empirical models to determine component performance across a range of geometries and
operating conditions (Atkins, 1999). Although the details of these models are generally
confidential, some published literature does exist on the development of such rating methods.
Stoneham et al (1980) proposed a number of generalised correlations to approximate evaporating

heat transfer coefficient for various refrigerants under non-standard conditions where only a few
variables are needed, although the errors are significant through using such a generalised
approach. These formulae were incorporated into a model for extending catalogue data by Trott
(1981). More recently Rabehl et al (1999) proposed a more comprehensive method of reducing
catalogue cooling coil data through correlating simple exchanger configurations and fluid
properties. It allow ed for more accurate extrapolation of coil performance data to other operating
conditions. Studies for condensers is sparse, although a paper by Holladay (1984) devised a
method of rating combined desuperheater/ condenser capacity7as a function of refrigerant and
water temperatures, and water flow rates. Kayansayan (1994) considered the effect of airside coil
configuration and correlated effects of various fin geometries on heat exchanger performance.
138
Following commercialisation of mixture refrigerants, some discussion on use of mixtures in
certain components, typically the compressor has been published. Murphy et al (1998) discussed
the potential errors when compressors are selected according to bubble, dew or “mid-point"’
temperatures. Errors manifest when a compressor is selected using a reference temperature
different to that for which it was rated. Hundy (2000) and Hundy and Vittal (2000) discussed the
same problem, but also offer a correction method to adjust the capacity' for a non-standard
reference temperature. This problem was also addressed in a practical sense (Jansen and Beks,
2000) where performance of compressors with a pure and mixture refrigerant was tested in a
calorimeter. When performance of the pure and mixture refrigerant w as compared using dew -,
bubble- and mid-point temperatures, the rated capacity' could be higher or lower than that o f the
pure refrigerant, depending on the conditions.
A theoretical study of Bobbo et al (1995) considered the choice of reference temperatures for all
system components and proposed that - at least for comparison purposes - the thermodynamic
mean temperature (defined as enthalpy divided by entropy) be used. The only discussion on heat
exchanger rating in respect to use of zeotropes w as a study by Ntalles (1997). The conventional
basic rating method was applied, but an ‘excess’ capacity introduced to account for the
differences resulting from the temperature glide. All rating values were based on the evaporator
outlet temperature.
Balancing
The conventional component balancing method is detailed in several textbooks; ASHRAE (1998),
Dossatt (1991), Trane (1988), and Stoecker and Jones (1982). The method described was
identical in each, and none specified whether they apply to pure only or both pure and mixture
refrigerants. Since most of these sources were published prior to the commercialisation o f mixture
refrigerants, it was assumed that they were intended for pure refrigerants only. Gosney and Suen
(2001) also detailed the conventional balancing method, and offered solutions to problems
involving system pressure losses. Discussion with Suen (2000) confirmed the methods (Gosney
and Suen, 2001) applied only to refrigerants without temperature glide. Only two further
references specifically covered balancing. One (Page, 1989) detailed the conventional method and
subsequently described how it could be applied to computer programmes and spreadsheets. Trott
(1981) did the same, although included formulas to calculate component performance at non
standard catalogue conditions. In all of these sources, balancing w?as based on a single saturation
“evaporating” and/or “condensing” temperature.
139
1.3.2 Implications of zeotropic refrigerants
With the search for alternative refrigerants, a large amount of studies related to mixture
refrigerants has been published. Whilst a minutia of publications addresses the topics under
consideration, findings on related subjects may provide additional insight into the behaviour of
zeotropic refrigerants that could impact on their effect on various stages in the component-system
chain. Therefore studies on heat exchanger analysis, heat transfer characteristics, composition
shift and overall system performance are of interest.
Analysis of heat exchangers with mixture refrigerants

The two main approaches for heat exchanger analysis are the logarithmic mean temperature
difference (InAT), and the Number of Transfer Units-Effectiveness (NTU-s) method. Both
approaches are based on several assumptions that are not necessarily valid when a phase-change
fluid is used. One is constant overall heat transfer coefficient ( U ) and specific heat capacity
( Cp) throughout the fluid path. Liu (1996) revised formulations for InAT based on these
variations. Additionally, NTU-s method assumes that phase-change fluids possess infinite
capacity' rates, and is not valid for zeotropes. Granyrd and Conklin (1990) derived new formulae
for NTU-s for phase-change with finite capacity rate, and later Conklin and Vineyard (1991)
utilised the revised relations to assess measurements on heat exchangers with high temperature
glide. The revised formulae showed good agreement with measurements when pressure loss in the
heat exchanger was low. These studies imply that use of either InAT or NTU-s conventional
methods will introduce error to analysis with zeotropes when applied to an entire heat exchanger,
unless they are suitably modified.
Much recent work on mixture refrigerants focussed on their positive or negative impacts on
system performance. One aspect in particular, is the potential optimisation of a system - through
improved heat exchanger performance - achieved with ‘'glide-matching'’. Mulroy et al (1994)
investigated this issue, and on the basis of a linearised temperature glide (and linear temperature
profile of a single-phase heat transfer fluid) found entropy generation could be reduced (i.e., by
obtaining smaller temperature differences). Measurements from a system using a tertiary mixture
(blended to linearise temperature glide) were presented to support the argument. Venkatarathnam
et al (1996) discussed the effect of mixture properties and specific glide characteristics in relation
to temperature profiles along heat exchangers. Later, Venkatarathnam and Murthy (1999)
considered the effect of changing refrigerant composition to show that using wide-glide binary
mixtures produced better glide matching.
140
Heat transfer in mixture refrigerants
It is known that the presence of two or more refrigerant components has implications on heat
transfer coefficient (/?), and hence exchanger performance. Earlier, Bivens and Yokozeki (1993)
explained the theoretical reduction of h , and attributed it to resistance of mass transfer of each
species as the mixture changes phase. A generalised formula was proposed for re-calculating h
of any mixture. Recently, Hewitt (1999) provided an overview' of research progress into h ,
including for mixture refrigerants. Based on review' studies by Wang and Chato (1995), Shin et al
(1997) and Stephan (1995), the consensus for the most accurate correlation for evaporation was
that of Jung and Radermacher (1993a and 1993b). Similarly, review' studies on condensation
were carried out by by Gaibel et al (1994) and Kenny et al (1994). However, a more recently
correlation by Shao and Granryd (1998, 2000) showed more favourable accuracy.
Composition shift
Other than the presence of a temperature glide and an effect on h , another characteristic of
zeotropic refrigerants is separation of species within the parts of the system where refrigerant
exists in two-phases. This results in localised ‘"hold-up” of certain species, thereby causing the
circulating composition to differ from the refrigerants' original charged composition resulting in a
possible departure from the intended operation. Measurements of variations of refrigerant
composition around a system were reported by Chen and Kruse (1995) and Sumida et al (1995),
and generalised models were developed in order to predict the composition shift. Chen and Kruse
(1995) found the accuracy of the model to be highly dependent upon the choice of void fraction
model, and the Hughmark correlation was shown to provide results that matched measurements
best.

Initially, the issue of increasing application of zeotropic refrigerants was presented, as was an
explanation of the flow of performance data through the component-system chain. In light of the
fundamental differences betw een the properties of conventionally used pure refrigerants, and the
recently commercialised zeotropic refrigerants, possible incompatibilities within the current
methods may exist. The review of the current practices within the component-system chain and
the behaviour particular to zeotropic refrigerants indicate the possibility of incompatibilities, and
therefore it is necessary to define an approach to examine and resolve these issues.
141
There are several key findings drawn from the literature review, as follows:
- Component rating and catalogue data. Presently no industry standards or manufacturers

information relate specifically to the characteristics of mixture refrigerants. A small number
of studies present options for converting component capacity data from one reference
temperature to another, where a glide is present. Whilst most standards require data to be
formatted in a specific manner, catalogue information appeared to be inconsistent in terms of
reference conditions. Use of correction factors appeared to be the most popular format for
data conversion. These observations highlight the need for a standardised approach,
especially with the introduction of refrigerants with temperature glide.
- System balancing. This conventional method is based on pure refrigerants only and accounts
for a limited number of parameters, and neglecting others such as pressure loss effects.
Considerable manipulation of data may be required to convert data from published reference
temperature to one consistent with other component data, so that system balancing is
possible. Evidently, there is a demand to improve the balancing technique and to ensure the
supply of consistent data for this purpose.
- Implications o f zeotropic refrigerants. The literature contained findings relating to a variety

of aspects associated with zeotropes. The heat exchanger analysis techniques such as NTU-s
were shown to provide accurate results when correctly revised for zeotropic refrigerants.
Different mixtures exhibit a range of glide profiles (variation of saturation temperature with
vapour quality), and this can have an impact on cycle efficiency. To optimise heat exchanger
performance, glide matching techniques are used to reduce approach temperature difference.
The studies on evaporation and condensing heat transfer coefficients showed degradation
with mixtures and provided information on favoured correlations for use in heat exchanger
models. Refrigerant composition shift was shown to occur in systems particularly where
mixtures have a large glide.
Little exploration of the influence of temperature glide on testing, rating and selection of
components was found. As such the understanding and recent developments in peripheral subjects
can provide understanding to assist with the current investigation. The review7of industry
practices for rating and catalogue information provided a base from where to start developing
techniques for accounting for the behaviour of zeotropic refrigerants.
142
This study links the two themes covered previously: zeotropic refrigerants and their integration
into the component-system chain stages. Presently, current methods employed within the
component-system chain are primarily based on the use o f pure refrigerants. Transition from a
pure to zeotropic refrigerants calls for revision of the current methods employed with respect to
the different behaviour exhibited by zeotropes.
The examination of the procedures within the various design stages found that the flow7of
information betw een stages is generally reliant on standard reference conditions for development
and interpretation of performance data. Since the difference betw een pure and mixture
refrigerants are primarily manifest as a temperature glide, the validity of currently defined
reference conditions is questioned. The stages within the component-system chain that rely on
choice of reference condition are: component performance testing/rating, catalogue data
presentation and selection, and system balancing.
Since each component-system stage is linked to the next, intuitively there must be consistency in
the reference conditions and interpretation methods applied throughout. Following an assessment
of each stage, associated methods need to be modified if necessary. Adoption of improved
methods should lead to improved accuracy in the design of a refrigerating system.
The general approach to developing modified or new7methods requires the performance of the
primary components to be evaluated, first in terms of pure refrigerants, and then re-evaluation by
introduction of the properties of zeotropic refrigerants. Initially adopting a simplistic analysis
should identify any differences in performance that are inconsistent with current rating, selection
and balancing practices. This will indicate where further work would be required to revise the
existing methods.
The general sequence of activities for investigating these issues is as follow s:

- Differences between pure and zeotropic refrigerants. The basic differences between pure
and zeotropic refrigerants are discussed, particularly in terms of inherent properties
associated with rating, selection and balancing activities. In addition, the primary
refrigeration system components are introduced, and general characterisation of their
performance is described.
143
- Further analysis o f zeotropic refrigerants in heat exchangers. The focus was on the
condenser and evaporator and their behaviour with and without refrigerants with temperature
glide. Initially a general analysis is used, and the findings examined further with detailed
modelling. This will help to conclude how reference states should be handled and leads to
new rating methods.
- Assessment and modification o f balancing technique. On the basis of the changes identified
for component rating methods, a modification to the existing balancing technique is proposed.
The procedure of the revised balancing technique is illustrated in a worked example, which
also acts to check whether the new methods achieve an improvement in accuracy.
Lastly, it is noted that the scope of the current study omits certain considerations to maintain
focus. In particular, only direct expansion systems are addressed and interconnecting components
such as pipew ork, valves, receivers and accumulators are ignored. In terms of mixtures, the
occurrence of composition shift and any influences associated with oils is neglected. The
components and systems under consideration are assumed to operate under steady conditions so
transient operation and influences of any controls are not under consideration.
144
SECTION 2: IMPLICATIONS ON COMPONENT PERFORMANCE
2.1 Initial consideration of impact of glide on component performance

The first activity is to clarify the basic differences in the properties of pure and zeotropic
refrigerants. The distinctive behaviour of zeotropic refrigerants is observed with respect to a
simple refrigeration cycle, as shown on the pressure-enthalpv diagrams in Figure 35. The cycle is
illustrated on both diagrams, where the refrigerant vapour enters the compressor at point (1),
discharged at (2), cools, condenses, and flows into the expansion device at (3), and is expanded
into the evaporator at (4), where it changes to vapour and reaches (1) again. This basic process
occurs with both refrigerant types. However, by following the cycle with respect to the lines of
constant temperature, a key difference is manifest. Whilst assuming isobaric flow' from (2) to (3)
and (4) to (1), the zeotrope in Figure 35(b) shows the condensation and evaporation process to
occur with vary ing temperature, and this is termed the temperature glide ( ATG ).
P p
r-o
-©
Figure 35: Pressure enthalpy diagram fo r (a) pure refrigerant, and (b) zeotropic re frigerant
Assuming the two cycles operate at identical pressures, the pure refrigerant changes phase at
constant temperature, or in other words, the saturated refrigerant temperature into the condenser
( T2s ) also known as the dew-point ( Tcdp), is equal to the saturated temperature at the outlet
( T3s ) also known as the bubble-point ( Tc bp). Similarly, with the evaporator the inlet temperature
(T4) equals the bubble-point ( Te hp), which is the same as the saturated temperature at the
evaporator outlet (Tu ) or dew-point ( Tedp). Conversely, the corresponding points for the
zeotrope show7a higher inlet temperature at the condenser compared to the outlet (i.e., Tc dp >
Tc bp), and a lower inlet evaporator temperature than its outlet (i.e., T4 < Te dp). Since for a pure
145
refrigerant Tdp - Tbp, the subscripts dp and bp are neglected and a single saturation
temperature is normally referred to.
Therefore, defining the performance of a component according to a single reference temperature

is appropriate for a pure refrigerant, but inexact for a zeotropic refrigerant without further
definition of the reference temperature. For the four main system components the performance
measure of interest are cooling and/or heating capacity. Therefore, it is important to identify how
component capacity (or definition of capacity ) is affected by selection o f reference temperature,
and to identify a means of choosing the proper reference temperature to apply consistently. With
regards to the condenser and evaporator, both are considered to be similar in operation and
therefore it was chosen to analyse them together in the follow ing sections.
2.2 Compressor
ISO 917 details methods for measuring performance and ISO 9309 suggests how performance
data are fitted to a non-functional polynomial equation to assist with generation of catalogue data.
The same formula is used for both evaporating capacity' ( Qe comp) (equation 76) and input power
( WcomP) (equation 77).
e 7comp
(76)
comp
(77)
where cx ... cK] are specific to Oecomp and Wcomp, and two reference temperatures, Te is
saturated temperature at suction pressure, and Tc is saturated temperature at discharge pressure.
Condensing capacity- of the compressor ( Qc comp) is normally assumed to be the sum of Qe comp
and Wcomp (although not specified in the standards). It is noted that for balancing purposes,
Q c,com P is the variable of interest, not Wcomp .
These equations were originated for pure refrigerants, so it is necessary to consider their
application with a zeotrope, and the associated impact of ATG on reference temperatures.
However, because equations (76) and (77) are simply curve-fits, the actual behaviour o f the
compressor is obscured. Therefore another approach was to examine the conventional functional
146
calculation method for compressor performance. Compressor performance is described by inlet
temperature and pressure of the superheated vapour ( and p x) and the outlet conditions ( T2
and p 2). Based on these conditions, the compressor evaporating capacity ( Q e ,comP ) 1S given by
equation (78).
e.comp Vr-fc-Hj (78)
where mr is mass flow of refrigerant 77, and H 4 are specific enthalpy of refrigerant exiting and
entering the evaporator respectively. Mass flow rate o f the refrigerant is defined in equation (79).
*” r = Tl Vo r VSw - P l (79)
where rjwl is volumetric efficiency of compressor, V^. is swept volume rate of compressor, and
Pj is density of superheated vapour entering the compressor. Compressor work, or the energy
applied to the refrigerant whilst passing through the compressor ( Wcomp) is expressed in equation
(80).
comp (80)
where H 2 is specific enthalpy of the superheated vapour exiting the compressor.
Since the capacity of the compressor is not reliant upon temperature differences to produce a
particular capacity, then the ATG itself has no fundamental influence on its performance.
Similarly, according to the standards ISO 917 and ISO 9309, there are no parameters within the
test or rating methods that apply to refrigerants in such a way that a ATG could affect the results.
However, in terms of equations (78), (79) and (80), reference points must be specified for
obtaining specific properties, implying that there must be some consideration as to which
reference temperatures are used to define the performance. The same is borne out with the
standards where the rating is provided as a function of saturated temperature. Thus a distinction
should be made between T* and . Since the compressor operation “sees” the evaporating and
condensing dew-point temperatures, these points are considered to be the preferred choice as the
reference. There should be no further considerations here with regards to temperatures and
therefore ATG.
2.3 Expansion device

A TEV was considered in this study since it is the most commonly used expansion device.
147
The rating procedure for TEVs given in ASHRAE 17 requires that refrigerant mass flow and
orifice aperture be used to determine its evaporating capacity ( Oe TEV) for any desired operating
condition. Calculation of capacity is given by equation (81), essentially being the product of
refrigerant mass flow rate and evaporating enthalpy difference.
O.
~ e.TEl’ (81)
where ^ is a correction factor dependent upon the amount of subcooling and superheating.
Evidently, determination of capacity7(from equation 81) does not rely directly on reference
temperature. The valve capacity is approximate due to the self-modulating behaviour within a
wide operating range. The effect of ATc does not impact on the selection methods except in
terms of the reference temperature. This is also the conclusion of Castle et al (1999) where this
aspect was investigated experimentally. Based on discussion of the choice of reference
temperature for the compressor, to be consistent the dew-point temperature corresponding to
evaporating and condensing pressure should be used for valve selection.
2.4 Condenser and evaporator

Determination of condenser and evaporator capacity7is currently obtained for reference conditions
detailed in standards EN 1117 and EN 1118, respectively. For presentation of data, the capacity
(Qc.cond or Qe.nup) *s re-calculated for other non-standard conditions according to some approach
temperature difference (equations 82 and 83).
(82)
where Q'ccond is capacity' at standard rating conditions, Arc/(,„) is the standard approach
temperature difference between condenser HTF and refrigerant inlet temperatures corresponding
to the saturated discharge pressure, and ATc(in) is the non-standard temperature difference.
(83)
where 0 ’e evap is the capacity7at standard rating conditions, ATîn) is the standard temperature
difference between HTF and refrigerant inlet temperature corresponding to the saturated suction
pressure, and ATe(in) is the approach temperature difference at non-standard conditions.
148
For both heat exchangers, the capacity is a function of a temperature difference betw een HTF and
the refrigerant. The existence of a ATc implies an impact on both AT[in) and AT{jn), compared
to the case of a pure refrigerant, and therefore equations (82) and (83) may no longer be
applicable to zeotropes. Therefore, clarification of the implications of zeotropic refrigerants is
required, and will be evaluated presently. In particular, the use of equations (82) and (83) will be
investigated in more detail, with the goal of identifying potential modifications to existing rating
and catalogue presentation methodologies.
HTF (in) Tf
Figure 36: Graphical representation o f (a) condenser, and (b) evaporator characteristic fo r
pure refrigerant (where B is constant)
2.4.1 Analysis of heat exchanger rating with pure refrigerants

Taking the general rating equations (82) and (83), the standard performance parameters
( 0 7 AT{in)) are termed the basic rating ( B ). By rearranging equations (82) and (83), and
substituting 0 7 AT'in) for B , provides the general formula for the basic rating of the condenser
(equation 84) and evaporator (equation 85) respectively.
(84)
r(in) * HTF (in)
=- Q (85)
1 HTF (in) 1 r(in)
where 7 ^ , . ^ and Tr(w) (i.e., Te and Tc ) are the inlet temperatures for the HTF and saturated
refrigerant respectively. The graphical interpretation of equation (84) and (85) are in Figure 36,
indicating the linear relationship between Q and 3sT(in). The use of the basic rating approach is
149
employed to generate catalogue data from performance tests and extrapolation of catalogue data
to application design conditions.
Thus, the foundation for the basic rating should be understood if it is to be adapted for zeotropic
refrigerants. If equations (84) and (85) are rearranged, they are found to have similarity- with the
NTU-e method (equation 86).
Q = s C mjn ■( T - T ) (86)
where the inlet temperatures of the hot (Thot(in)) and cold ( Tcold(m)) fluids relate directly to the
inlet temperatures of the refrigerant and HTF, where B equates to the term s * C mn, being the
product of exchanger effectiveness and minimum capacity’ rate of the two fluid streams (equation
87).
B = s -C mm (87)
For a single-phase fluid, capacity’ rate is the product of the mass flow rate ( m ) and its specific
heat ( C p ) as equation (88).
C -m C p (88)
When equation (88) is applied to fluid undergoing changing phase Cp is infinite, thus C is
infinite. In equation (87), C mm refers to the fluid stream that possesses the lower capacity rate of
the two. Since C is infinite for a pure refrigerant, Cnnn must belong to the HTF ( CHTF).
Further, for a heat exchanger with fixed HTF mass flow rate ( mHTF ), Cmin is constant. The
exchanger effectiveness term in equation (87) is normally presented as a function of the non-
dimensional number of transfer units ( N TU ) (equation 89) and capacity’ rate ratio ( Rc ), being the
ratio of Cm!n and the maximum capacity’ rate Cmax (equation 90), i.e., that of the pure
refrigerant.
TJ • A
(89)
tmn
Rt = (90)
max
Consequently R a —*■0, which reduces the calculation of effectiveness to equation (91).
150
s - 1- exp (~N tu ) (91)
where A is the exchanger surface area and U is the overall heat transfer coefficient. If U is
assumed to be constant regardless of T r(in) , N tu will be constant and therefore so will s ; thus it
follows that the supposition of constant basic rating is valid.
r(in)
HTF (in ) HTF (in) HTF (in)
distance distance distance
Figure 37: Temperature profiles o f counter-flow condenser fo r reducing condensing

temperature with pure refrigerant
A graphical representation of the heat exchanger capacity with saturated refrigerant temperature
is shown in Figure 36. It is seen that reducing the temperature difference by shifting Tr(w} (i.e.,
Te or Tc) reduces capacity until the point at which THTF(m) - Tr[m) and 0 = 0 . The relationship
between capacity and approach temperatures may also be shown through the visualisation of
refrigerant and HTF temperature profiles. Figure 37 illustrates the effect of a lowering
condensing temperature in a counter-flow heat exchanger in three stages, with the area between
the profiles representing the capacity. To the left is the case of a wide AT{in) where a large
quantity of heat is transferred throughout the entire distance of the heat exchanger. Next, the
central diagram show s a medium AT{in) where a reduced quantity of heat is transferred. Lastly,
to the right, nearly equal approach temperatures give miniscule AT(in) where negligible heat is
transferred. Whilst the shape of the refrigerant temperature profile remains constant over the
exchanger distance, the shape of the HTF converges towards the refrigerant logarithmically
(assuming U , A and CpHTF remain constant).
2.4.2 Analysis of heat exchanger for refrigerants with temperature glide

The NTU-e basis was used to investigate the influence of ATG on exchanger rating. In the first
place the definition of B is considered, latterly the approach is used to address overall exchanger
151
behaviour. Each of the parameters in equation (86) was considered with respect to the properties
of a zeotrope. An overview of equations (86) to (91) identifies the quantification of the
refrigerants Cp as the main difference between pure and zeotropic mixtures, since the existence
of a refrigerant temperature rise/fall during phase-change implies a finite Cp . If Cp is finite,
then equation (88) suggests that C of a zeotrope is also finite, and therefore Cmjn cannot
automatically be assumed to belong to the HTF and B becomes dependent on Cmin .
Determination of Cp was achieved through transposition of exchanger energy balances. Using
the conventional formula for sensible heat gain of a fluid under steady flow gives equation (92).
Q = m Cp-(T(out) - T (in)) (92)
Equation (93) represents the conventional heat balance for steady flow undergoing phase-change.
Q= (93)
Combining equations (92) and (93) provides a definition o f the “apparent'’ specific heat of the
refrigerant ( Cpr ) (equation 94).
(94)
(out) (in)
If equation (94) is applied to a pure refrigerant, then T(out) = T(in) and therefore Cp'r = oo.
However, for a zeotrope undergoing phase change where T{out) - T(m) = ATG, then Cp'r < oo.
Further, substituting equation (94) into equation (88) provides the “apparent” capacity rate of the
refrigerant (C'r ) (equation 95).
(95)
•* (out) * (in)
Similarly, if equation (95) is applied to a pure refrigerant, C'r = oo. It is noted that equations (94)
and (95) only provide mean values, representing a linear glide with respect to enthalpy or quality
change. Since most mixtures exhibit a non-linear glide, Cp’ and C'r are more accurately defined
with respect to local vapour quality; equations (96) and (97), respectively.
(96)
‘K - ' f
152
Equation (95) demonstrates that a greater ATG gives low Cp' and C ' . Now that C ' < oo, the
capacity rate ratio (equation 90) must be greater than zero, and therefore exchanger effectiveness
is dependent on ; thus equation (91) is no longer valid. With > 0, effectiveness is
dependent on exchanger geometry, and for parallel-flow and counter-flow (for concentric tube),
Incropera and DeWitt (1990) offer equations (98) and (99) respectively, and for cross-flow,
equation (100).
l-exp(-A rr„ -(l + ^ ) )

s - (98)
1+ R,
l - e s p ( - N TU- ( l - R d ))
(99)
1- Rd •e x p ( - i V - a -«< ;))
r N TU 022 0.78
£ - 1 - exp ■( « p ( ~ R cC ( 100)
R±
parallel
i
counter
cross
O
fa
<u 0.5 ? 0.5
U
cCOi parallel
JSo
x counter
W
cross
0.0 0.0
0.0 0.2 0.4 0.6 0.8 1.0 0.0 0.2 0.4 0.6 0.8 1.0
Zeotrope apparent capaciy rate (kW K’1) Zeotrope apparent capacity' rate (kW K~ )
Figure 38: Effect o f apparent capacity rate on Figure 39: Effect o f apparent capacity rate on
effectiveness basic rating
Interpretation of the revised NTU-e formulations revealed that with different ATG, C ' can be
higher or lower than CHTF, meaning that Cmin could be either C'r or CHTF . More importantly,
different ATG will affect both N ru and Rc , where a greater ATa lessens N TU and R^,.
To illustrate the differences between a pure and zeotropic refrigerant, effectiveness is shown as a
function of C ' for the three exchanger geometries (from equations 98, 99 and 100). The results
are shown in Figure 38, where calculations assumed linear glide and constant U and CHTF = 1
kW K 1. With a smaller C ' (i.e. larger ATG), 6 always tends towards unity , where the counter
flow geometry converges on s = 1 first and the parallel has s < 1 until C ' —►0 kW K"1.
Introducing C ' = Cmw to the values from Figure 38 (as equation 87), Figure 39 shcnvs how7the
basic rating varies with C ' due to ATG, when CHTF = 1 kW K"1. Whilst with a pure refrigerant
B is constant (due to C ' = Cmax), there is a wide variation for a zeotropic refrigerant, depending
upon the geometry. As with effectiveness, the counter-flow exchanger gives the highest B and
parallel-flow the least.
2.0 1.0
- 0.8
- 0.6
oo - 0.4
0.5 Cirin
. . . - 0.2
basic ralng
effectiveness
0.0 -i----- —----- 1- 0.0
0 5 10 15
Refrigerant temperature glide (K)
Figure 40: Effect o f temperature glide on Cm,„. e and B in counter-flow exchanger
Figure 40 show s explicitly, the effect of ATG size on Cmw and s and subsequently, B . The
dashed line indicates = 1; to the left C'r > CHTF and C ' < CHTF to the right. When the glide
is small, C'r is relatively large, and because CHTF is constant regardless of glide Cmm is also
constant.
As the glide increases C ' gradually reduces from infinity until at ATG ~ 6 K, it falls below
CHTF , and the trend of Cmm therefore begins to reduce. With zero glide, £ is high due to =
154
0, but as C ' —> CHTF, s also reduces until Rc = I, and then as C ' falls below CHTF, e
increases. The combined effect o f s x shows a smooth reduction in B with increasing
glide. The implication is that for a condenser or evaporator using a zeotrope, the basic rating is
(theoretically) Iow?er than if a pure refrigerant was used. Since B is dependent upon ATG, there
are additional implications since it is known that the glide tends to vary with saturation pressure
(discussed later).
- - C'r = 0 0 - - rim)
S 'r
C -T HTF [ HTF(out)
distance
Figure 41: Temperature profiles fo r counter-flow condenser
In analysing the characteristics of a heat exchanger using a zeotrope, the behaviour of the
temperature profiles provides insight. Previously, it was shown that with a pure refrigerant, heat
exchanger characteristics meant that when A7'(m) —►0, then O —►0 (Figure 36). With a
zeotropic refrigerant, the analysis is complicated due the existence of a ATc .
IT ,I i >i A
rim )
k ----- THin) Tnm)
k ^ Tnm) •te r a lle j r{in)
THTF {in) 1HTF (in) 1HTF (in)

distance distance distance
Figure 42: Temperature profiles ofparallel- and counter-flow condenser fo r reducing

condensing temperature with glide
155
Possible refrigerant and HTF temperature profiles are illustrated in Figure 41, which indicates
three categories for the temperature gradients depending upon size o f glide: (i) C ' > CHTF for
small ATg , (ii) C'r = CHTF for medium ATG, and (iii) C'r < for large ATc , for the
example of a counter-flow condenser. The temperature profile for a pure refrigerant ( C ' = 00 ) is
also included for comparison. It is the latter category (iii) that is of most interest, since it implies
a greater departure from constant B , and it contradicts the conventional exchanger characteristic
in that it suggests AT{in) must be greater than zero when THTF{m) = THTF(out), i.e., 0 = 0 .
This matter is investigated further by establishing the consequence of having incrementally lower
Tr(w), as in Figure 37 for pure refrigerants. For a zeotrope, Figure 42 illustrates a reduction of
Tr(tn) and hence AT{in) (for both parallel and counter-flow type condensers), assuming a fixed
ATg and . Initially (left) the wide AT(jn) results in a large capacity and correspondingly
a large HTF temperature rise. Lowering Tr(in) reduces capacity and correspondingly lessens the
HTF temperature rise, but THTF(out) is still above Tr[out). Tr{m) is reduced further, but must cease
when AT(oul) = 0 K, and this corresponds to a minimum approach temperature difference
( ) ^r(out) cannot be less than the temperature of the HTF stream at any point w ithin the
exchanger. In this final case fluid stream temperatures converge, Q should approach zero;
although this seems paradoxical since some temperature difference between the fluid streams
exists.
HTF (in)
condenser distance evaporator distance
Figure 43: Minimum approach temperature differences fo r condenser and evaporator
156
On the basis that the capacity should be zero when AT-out) = 0 K, then for the phase-change
region only, ATmin(inj must be equal to ATG . This situation applies to both parallel- and counter-
flow heat exchangers, and although not shown, to cross-flow exchangers as well. Essentially, the
difference between the HTF inlet and refrigerant inlet temperatures is the offset due to the ATG,
as shown graphically in Figure 43. On this basis, the graphical representation of the condenser
and evaporator characteristic in Figure 36 is modified by shifting the THTF(w) for the amount of
the ATg (Figure 44). Presently, the basic rating is assumed constant - resulting in the linear
relationship between O and AT(lw) - although the earlier discussion implied this may not be
correct, and therefore demands further consideration.
c,cond z~e,evap
H T F (in)
T
1c
^ F jn in (in ) — ejnin(m )
Figure 44: Graphical representation o f (a) condenser and (b) evaporator characteristic with
zeotropic re frigerant
2.4.3 Implications of temperature glide behaviour

The prior analysis assumed linear temperature glide with respect to vapour quality and constant
overall heat transfer coefficient. Observation of temperature glides of real mixtures often deviate
from this linear form, and similarly U is also known to vary. Initially, characteristics of the
temperature glide were addressed, and in particular tw o properties were identified for further
investigation: variation of ATG with saturation pressure, and “shape'’ of ATG with respect to
vapour quality.
Properties of the temperature glide

An appreciation of the glide behaviour of zeotropes was gained through examining the ATG
properties of various commercialised refrigerants. Details of zeotropic refrigerants w?ere taken

from UNEP (2002), and properties calculated using the subroutines of Refprop (Lemmon et al,
157
2002). A list of 51 refrigerants and their compositions and selected properties are listed in Table
31 of Appendix 2.
First, total ATG (i.e., over the entire phase envelope) was calculated for a range of pressures that
corresponded to saturated dew-point temperatures from -40°C to +60°C. The results are
presented in Figure 45 (a); each refrigerant has not been identified individually because the
purpose of the graph is to illustrate the general trend of change in ATG with saturation
temperature. Towards the lower temperature range some glides reach up to 12 K, but as
saturation temperature approaches the critical point ATG —> 0.
10 10
concave
8
&
convex
l
4
0.1 0
-40 -20 0 20 40 60 0 0.2 0.4 0.6 0.8 1
Saturation temperature at dew-point (CC ) Vapour quality, x (-)
Figure 45: Variation o f temperature glide (a) with saturation temperature, and (b) with vapour
quality at atmospheric pressure fo r commercial refrigerant blends
About 10% of the mixtures exhibit a ATG that does not follow this trend (although they will
eventually reach zero at the critical point), and these are shown as dashed lines in Figure 45 (a).
These fluids were identified to be two sets of mixtures. The first set was two R32/R125 blends,
which have small and relatively constant ATG (-0.5 K and -1.5 K) over the temperature range,
but peak in size around 0 - +20°C. The second set was three R134a/R142b blends that had a
trend contrary to the others where ATG tended to increase with higher saturation temperature,
and their glide also ranged around 0.5 K to 1.5 K. Although the relatively wide range of these
blends’ partial pressures would lead to a large ATG, these observations suggest that the mixtures
exhibit azeotropy at compositions close to those of the refrigerants composition. They are
therefore not considered to be “true” zeotropes and are subsequently neglected from the following
analysis.
158
Given the generally steady trend in ATc with saturation pressure seen with most refrigerants, it is
useful to characterise it. The variation in ATc was found to follow saturation pressure
exponentially (equation 101).
& t g (P s ) = c g\ -exp(cG2 - p s ) (101)
where p s is the saturation pressure, and cG1 and cG2 are constants. This exponential trend
provided a good fit with the refrigerant data. In correlating the data, it was found that cG] was
comparable to ATG except at very low7pressures. After evaluating different pressures, the closest
fit included ATG at standard atmospheric pressure, thus cGX = ATG(p alm), as Figure 46 (a).
12 T OE+OO t -
DO
/O -5E-04 -
e o
3»
ao
o
0 4 8 12 0 4 8 12
Temperature glide at p atm (K) Temperature glide at p atm (K)
Figure 46: Correlation fo r constants (a) cG1 and (b) cG2 for equation (101)
Similarly, Figure 46 (b) also shows an approximate linear correlation between the constant cG2
and ATG(p alm), as equation (102); since the purpose of this exercise is to obtain a general
appreciation of ATG characteristics, this was accepted despite the scatter.
C G2 — a \ ' ATg ( p atm) + a 2 ( 102)
Lastly, by including the relationship between cGl and cG2 to ATG{pam) , equation (101) may be
rewTitten to characterise the change in temperature glide with saturation pressure, based on a
chosen reference ATG(p atm) (equation 103).
&Tg t p s) = ATg (p atm) •exp ( [ax • ATG(p a[m) + a2] - p s ) (103)
159
where ax = 0.0000143, and a 2 = - 0.00036.
The second property is the “shape” of the glide, or the variation in local saturation temperature
with vapour quality as the zeotrope moves betw een dew - and bubble-points. Figure 45(b) show s
the variety of glide shapes exhibited by the different refrigerant blends, which include convex,
linear and concave behaviour.
For all binary mixtures (at any pressure), those containing a greater proportion o f the higher
pressure component leads to a concave glide shape, whilst equal mass fractions tend towards
linearity. Apart from possessing smaller ATC, a mixture of components with similar saturation
vapour pressures tend to a linear shape over a wider range of compositions. The variation of
saturation temperature (at constant pressure) with vapour quality (Tx ) was found to follow^ the
form of equation (104), by employing a power-law’ relationship.
T A P ,) = T ,„ (p ,)-A T a (p t ) - ( l - x ) " a (104)
where T+ is the saturated dew-point temperature, x is the local vapour quality, and nG is an
index dictating the shape of the glide for a given refrigerant: nG > 1 produces a concave shape,
nG = 1 is linear, and nG < 1 for a convex glide, see Figure 45(b).
o2 'O2
s:
X
TuC5
V
3 1
o
0
-40 -20 0 20 40 60 500 1000 1500 2000 2500
Saturation tenperature at dew-poirt (°C) Saturation pressure at dew-point (kPa)
Figure 47: Variation o f maximum and minimum glide index over (a) saturation temperature,
and (b) saturation pressure
160
nG was determined for the different refrigerants over the range of -40°C to +60°C, and was
found to vary between -0.5 to -2.5. The maximum and minimum limits of nG for the entire set
of refrigerants over this temperature range are shown in Figure 47(a), and also for saturation
pressures between -100 kPa to -2500 kPa in Figure 47(b). So that the remaining refrigerants’
nG are located between the maximum and minimum limits indicated. It is seen in Figure 47 that
as the saturation condition approached critical point, the maximum and minimum values of nG
—►1, and this demonstrates that the shape of the glide tends to linearise correspondingly. This
implies that any complications associated the temperature glide is relatively greater for the
evaporators operating at a lower temperature due to a larger glide size expected as well as having
a non-linear glide shape.
Given the trend of nG indicated in Figure 47(a), it was correlated with temperature and for
consistency with equation (104), Tdp was chosen as the correlating variable. An appropriate
value of iiG can be estimated from equation (105).
na = a\ -TdP{Ps) + a 2 (105)
where for a given refrigerant, ax is between -0.015 and 0.002, and a 2 is between 2.5 and 0.6,
depending upon the desired shape, although correct selection of the constants will cause nG = 1
when Tdp = Tcr
Influence of glide properties on basic rating

With reference to the discussion in Section 2.4.2, basic rating for a pure refrigerant is constant,
but when a zeotrope is used it is affected by the size of the glide (e.g., Figure 40). A variation of
ATg affects C ', and therefore B . Previous sections showed that ATG varies with saturation
temperature, which implied additional variation of B over a range of evaporating or condensing

temperatures. An example of the influence of the variable ATG is shown in Figure 48, w'hich was
calculated for a counter-flow7condenser. The filled data-points refer to the variable ATG
(equation 103), and the hollow data-points were calculated for a fixed ATG regardless of
saturation pressure.
For a pure refrigerant ( ATG(p atm) = 0 K), B is constant under all conditions. For a refrigerant
with a small glide (say, ATG(p atm) = 2 K), there was only a small variation of B with saturation
161
temperature and whether the variation of glide is accounted for or not has little effect on B .
However, in the case of a large glide (say, ATG(p atm) = 8 K), it is seen that a significant
difference in B can occur when the variation of glide size with saturation temperature is
included. This trend amplifies the importance of accounting for a variable ATG, when the glide is
large. For both variable and fixed glide cases, B reduces as Tc(ln) —►THTF[m) (i.e., ATc{in) gets
smaller).
0.8
atg( ^ ) = ok
^c(Palm ) = 2 K
A7’c (pato) = 5K
A7’G(pfl#n) = 8K
o
35 40 45 50
Inlet saturation temperature at dew-point (°C)
Figure 48: Change in condenser basic rating with inlet temperature fo r fixed (hollow data
point) and variable (filled data point) temperature glide
Particular influence on evaporator

The condenser was used as an example because it exhibits complete phase-change, as opposed to
the evaporator, where if the refrigerant enters the evaporator at vapour quality between 0 and 1,
the temperature glide is bisected. As such, the local refrigerant temperature is between bubble-
and dew -point, and therefore the ATG experienced by the evaporator is smaller than the ATG for
a complete phase-change. The temperature at w hich liquid refrigerant enters the expansion device
{TTW{in)) determines the evaporator inlet quality ( x e(m)), and therefore the size of the partial
temperature glide, i.e., ATGe = Te dp - Te(m). Earlier it was shown that the size of ATGe had
implications on B , and it therefore follows that B must be dependent upon TTEV(m). This is
illustrated in Figure 49(a) where different TTEl,(m) are shown on a pressure-enthalpy diagram, and
the corresponding evaporating inlet temperatures are shown for a fixed evaporator pressure.
162
Considering the basic rating equation (85), the evaporator characteristic should be re-plotted, and
this is shown in Figure 49(b) where a different curve is presented for each .
The implications of changing TTEV{m) can also be evaluated in terms of the heat exchanger
analysis. A higher TTEV{in) and subsequent smaller ATG e will result in higher C ' ultimately
changing 5 . It is also noted that because local refrigerant temperature rarely changes linearly
with x , the variation in B^,ap will depend upon the glide shape. These implications were
addressed by evaluating evaporator characteristics with the different refrigerants from UNEP
( 2002 ).
It was noted that whilst a change in x e(in) affects ATGe, refrigerant heat transfer coefficient (h r )
is also reliant upon vapour quality, and therefore, a change in TTEV(in) may also influence U ;
both of these ( ATGe and U ) affect evaporator capacity. Whilst a higher TTEV{m) will raise Te{in)
and therefore reduce Oe, the shift in xe(in) implies a higher U , because local hr typically
increases with x resulting in larger Oe . The degree of variation Oe depends on the relative
influence of ATGe and U . In fact, a shift in x e(in) also influences pressure loss along the
refrigerant path which impacts on the overall temperature glide, but this aspect was neglected for
the time being.
Figure 49: Effect o f different expansion device inlet temperature on (a) evaporator inlet
temperature, and (b) on evaporator characteristic
163
For the analysis, variation in U was accounted for by simply adjusting a mean value ( U )
according to x e(in). The evaporator model (detailed later in Section 3.2) was used to obtain local
values of U , which were used to calculate the area-weighted U for a range of xe(m) under
different sets of input conditions, geometries and refrigerants. Most situations showed an upward
trend in U with xe(w) . For the purposes of this exercise, a simple linear formula was used to
represent the variations in U (equation 106).
= <106)
where U (0) is the mean overall heat transfer coefficient when x e(m) = 0. The value of a is
largely dependent upon the input condition, geometry and refrigerant, but was typically found to
be between 0.05 - 0.1, and from 0 to 0.3 in extreme cases.
0.25
0 U = const
^ °-2° ij a U - variable
Mb (N oVO
W j A
A A
I 0151
Op I
1 0.10 j * •(«) = 20°C, U = const
A T t e v (m) = k = const - 10%

m 0.05 J o cP
0 ^TEV(m) ~ 20°C, U = var
TEV (m) = 60°C, U = var

-20° b
0.00
0 4 8 12
0 4 8 12
Full temperature glide at 0°C (K) Full temperature glide at 0°C (K)
Figure 50: Variation in evaporator basic Figure 51: Variation o f evaporator capacity
rating with glide size at different Ttev(m with glide size due to change in TEV inlet
temperature from 20°C to 60°C
The NTU-e method detailed above was used to calculate the capacity of a counter-flow
evaporator using various refrigerants. Oe at a fixed Te dp = 0°C was evaluated twice, for TTEV(m)
at 20°C and 60°C. Other inputs included CHTF = 1 kW K'1, THTFim) = +5°C, and U (0) = 200
W m'2 K'1, and a in equation (106) was fixed at 0 or 0.3. Some results are shown in Figure 50,
where is plotted against the full ATG at 0°C for each refrigerant. For any TTEV (.in) , there is
a general trend of reducing as the full ATG gets larger. The scatter is partly due to the
164
variation of refrigerants' glide shape, but it is also influenced by the gradient of the bubble-line
curve, which dictates the change o f x e(in) with TTEV[in).
When U is constant (filled data-points), there is a minimal difference between B âp at TTEV{m) =
20°C and 60°C at smaller ATG, but an increasing difference occurs with larger ATG . With
Ttei' w = 20°C, the variation in B ^ ap with ATG is smooth, whilst greater scatter is seen for
TTEV(W) = 60°C. This is attributed to the increasing influence of the gradient of the bubble-line,
as this moves the point of evaporator glide intersection further aw ay from Te bp thereby bringing
out the specific glide characteristics of each refrigerant. When the contribution of heat transfer
coefficient is introduced by varying U with xe(l/j) (hollowr data-points), the difference between
B ^ at the two TTEV(in) is large throughout the range of ATG . Again, when TTEV(Jn) = 20°C, the
change in B over the range of refrigerants is relatively smooth, and follow s the corresponding
data for constant U . However, when TTEr(m) = 60°C, B is on average consistently greater
than TTEV{w) = 20°C across all refrigerants, although significantly more scatter is seen that all
other cases. This is to be expected given that the stronger dependence of the gradient of the
bubble-line does not only affect the resulting ATG e, but xe(in) and therefore sensitivity to U as
well.
The data is also presented in Figure 51 in terms of the relative difference between Oe over the
two TTEV(in), for both constant and variable U . With constant U , Oe for TTEV(m) = 20°C and
60°C is seen to reduce by about 15% as the refrigerants' ATG —» 10 K, whereas a variable U
initially produces a 10% higher Oe at ATG ~ 0 K, but then eventually show s a 5% reduction in
Oe . Again, the scatter associated with the data is a result of the refrigerants’ different nG and
bubble-line curves.
Several important conclusions can be drawn from the discussion above. Primarily, evaporator
characterisation is dependent upon the inlet condition o f the refrigerant, and for any specific
refrigerant, this is dictated by the inlet temperature to the expansion device. A refrigerant with a
larger temperature glide typically exhibits a greater variation in Oe for a given change in
TTEV(w) , although the shape of the glide, and the gradient of the bubble-line have a some
165
influence. Assuming a fixed U , a higher TTEV(w) normally leads to a reduction of Oe on account
of the approach temperature difference (i.e., THTF{m) - Te{m)) being smaller. However, if U
increases with higher TTEV{m), the variation of Oe due to changing ATa is offset to an extent,
and if U has as strong dependency on x , this offset may negate any reduction in Oe . In fact, for
a refrigerant with a given set of glide characteristics used within a certain evaporator the
behaviour of U (xe(in)) may be such that Oe becomes effectively independent of TTEV(in).
2.5 Findings relevant to component-system chain and additional considerations

A number of findings were made, which confirmed the necessity to clarify suitable methods used
within the component-system chain when using zeotropic refrigerants. The foregoing review' of
the main system components with respect to the use of zeotropic refrigerants provided an
indication of areas where component capacity and performance rating differ compared to a pure
refrigerant.
Performance testing, rating and presentation of data for compressor and TEV requires some
reference value to be specified that correspond to the inlet and outlet state, and because of the
ATg , it follow s that the choice of reference state must be redefined when a zeotrope is used. Due
to the ATg , the conventional definition of approach temperature difference for evaporator and
condenser is not suitable. The size of the temperature glide changes with saturation temperature
which affects the basic rating and the capacity, meaning that these parameters must also be
included in the component characterisation. In the case of the evaporator, a change in TTEV{in) and
therefore inlet qualify' changes ATG which also influences the approach temperature difference.
Consequently, an alternative rating method is needed and definitions of reference states should be
identified.
A key aspect of the study is to establish suitable reference states for the rating and presentation of
data of each component, and the subsequent system balancing. When considering each component
in isolation, it was generally concluded that the dew-point was the most suitable choice. How ever,
when components are matched during the balancing procedure, it is important to ensure that
consistent reference states are used as each must “see” the same reference state as the adjoining
component. In the conventional technique, evaporating and/or condensing temperatures that
correspond to the saturated evaporating/ condensing pressures are used. Neglecting temperature
166
and pressure changes within interconnecting piping and ancillary components, this provides a
consistent basis for reference states of a pure refrigerant.
In terms of a zeotrope, saturated evaporating and condensing temperatures may be associated

with the dew-point, bubble-point or somewhere between, such as the mid-point temperature (half
way between bubble- and dew-points). Whilst the bubble-point would be suitable for the
condenser, it is not directly experienced by the evaporator. Similarly, because the saturated
temperature at evaporator inlet is variable, the choice of mid-point would be inappropriate
because it would change as the performance of the evaporator changed. The only reference state
that would remain consistent for all components is the dew-point, and therefore this is adopted
hereon.
Several parameters have so far been neglected which individually or in combination, may impact
on the interpretation of exchanger performance.
- Superheat, desuperheat and subcooling. Inclusion of sensible heat changes to the refrigerant
was omitted, despite being nearly always present in real condensers and evaporators. These
aspects may contribute to changes in the heat exchanger capacity and thus introduce a
departure from conclusions from the NTU-e analysis which assumed phase-change only.
- Pressure loss. This is of significance in the case of phase-change fluids since a sufficiently
high pressure loss will cause a marked change in saturation temperature through the change
in saturation pressure. The effect on temperature profiles when using zeotropic refrigerants
are such that a pressure loss in the condenser would effectively increase the temperature
glide, but reduce it in the evaporator. This would affect heat exchanger performance.
The most appropriate approach to evaluate the cumulative effect is considered to be combining
them in the form of a heat exchanger simulation.
167
SECTION 3: CHARACTERISATION OF COMPONENTS WITH ZEOTROPES
3.1 Approach for improved characterisation

Initially, it was proposed to use dew-point temperatures as the new reference state. However, it
was also noted that this conclusion was based on simple analysis of evaporator and condenser
performance, with certain parameters omitted. Therefore it was deemed necessary to further
evaluate their performance using a more detailed modelling approach. The simulation results will
help to clarify the appropriate definition of reference states, and in addition, identify the
functional dependency of capacity on relevant parameters that can be used for the rating method.
Further, performance data is required in a later section for the balancing exercise, so these models
can be used to generate fictitious data. This is also needed for the compressor, so a basic model
was similarly produced to generate fictitious performance data. 12
The following sections first detail the performance models for the compressor and then the
condenser and evaporator, where a generic heat exchanger model is described, and correlations
specific to either condenser or evaporator are given separately. The models are then used to
investigate the influence of various aspects such as glide shape, pressure loss and sensible heat
transfer, and from this conclusions are drawn relating to definition of reference states. Lastly,
possible options for revised characterisation methods are presented.
3.2 Simulation of compressor performance

Generation of compressor performance data requires information on discharge superheat, and
since conventional compressor catalogue does not normally provide this, using a simulation
approach is the only means by which it can be obtained. It is important to ensure that the data are
based on the newly established reference states for zeotropic refrigerants.
3.2.1 General modelling approach

A semi-empirical model was chosen representing a conventional single-cylinder reciprocating
machine. Existing catalogue data was employed to generate the required coefficients. It must be
noted that it was not necessary to use a particularly accurate approach provided the results were
conceptually sound and consistent with normal compressor operation.
The compressor, condenser and evaporator models were also utilised by cycle models in Part 1.
168
Relevant inputs for the model are swept volume flow rate, saturated evaporating temperature,
suction pressure, saturated condensing temperature, discharge pressure, evaporator superheat,
and suction superheat; saturated condensing and evaporating temperatures correspond to dew-
point. The output data required for system balancing are evaporating capacity, compressor power
or condensing capacity, and discharge superheat.
The model programme was written in Visual Basic (VB), and where necessary, refrigerant
property data was called from the subroutines within the Refprop database (Lemmon et al, 2002).
3.2.2 Semi-empirical model

The total evaporating capacity o f the compressor is obtained from enthalpy balance, equation
(107).
O
x e+comp ^SH ■ V r o l P\,SUCt ’ (77] 5 uct 7 /4) (107)
where p Xsuct is suction density7and ( H ] suct - H 4) is specific enthalpy difference between the
evaporator inlet and the compressor suction. rjvol is the volumetric efficiency and was calculated
with a modified formula based on Gosney (1982) (equation 108). The original formula was
intended for a single refrigerant, so the initial constant was adapted to account for differences
resulting from the use of other refrigerants. This was achieved by correlating the initial constant
with the molecular mass ( M ) and saturation pressure at 0°C ( p ()) of the different refrigerants.
f _ r \ h'r
Pi Pi
n.oi = <?, + c2 •M R-cvl R c \I + c. -1 + (c4 + c5 - p 0) (108)
\P \ J \P \ J
where is the “apparent” clearance volume ratio, p } and p 2 are suction and discharge
pressures respectively, and y is the isentropic index, defined as the ratio o f specific heats at
suction conditions (CpXsuct / Cvlsuct). The constant cx is a function of the specific compressor
geometry7and the remaining c2 ... c5 relate to the refrigerant type; all were obtained from
correlating catalogue data.
Determination of compressor power ( Wcomp) was also based on Gosney (1982), which adopts the
ideal mean effective pressure term ( p ime). The power was calculated as a linear function of
volumetric flow and p ime, and includes a constant value to account for frictional effects ( Wfr)
(equation 109).
169
W c o m p = W fr + b • P ,m e ' ' 1,-al )
(109)
where b is an empirical constant, and the product p ime • (Kw • ijvol) is the ideal compressor
work; p ime is from equation (110).
f V ''2
r r P i ' ( 110)
Pime 'P i • 1+ Rcvl ~ Rcvl
7 -1 \ P 1 )
The condensing capacity, Qccomp, was calculated as the sum of Qe comp and Wccomp '
Determination of the discharge superheat was achieved by calculating discharge temperature

from local enthalpy and pressure, using the property routines within Refprop. Whilst discharge
pressure is known, determination of discharge enthalpy (H 2disch) was based on an energy balance
across the compressor, which assumed 95% contributes to the heat of compression and 5% loss
to the surroundings (equation 111).
0.95 • W,comp
^ 2,disch H \ suet T> (111)
^ sw - P r o ! ' P i
where the lower term represents refrigerant mass flow rate. Discharge superheat was based on the
difference between discharge temperature and dew -point temperature corresponding to discharge
pressure, i.e., A = T 2Jisch - T 2 dp.
3.2.3 Determination of constants

The constants in equations (108) and (109) were obtained from catalogue data, and in addition,
because the actual clearance ratio was not given in the product information, an “apparent” value
R ^ i, was also determined from the catalogue data.
Information was used for several different compressors from manufacturers’ catalogue (Bitzer,
undated). For each compressor, capacities and power was given for several different refrigerants:
R22, R134a, R404A and R407C. For equation (108), the constant cx ranged from 5 to 40 as
compressor size increased. The best-fit for the remaining constants w7ere: c2 = -0.093, c3 =
0.184, c4 = 0.785, and c5 = 0.00027. For equation (109), b = 1.24, and Wfr varied according
to compressor choice, but generally represented 5 - 10% of Wcomp depending on the operating
170
conditions. Lastly, R^, was approximated (since it is not provides in manufacturers data) by
adjusting its value in equation (108) and equation (110) until the highest regression constant (r2)
was achieved for all refrigerants in any one compressor model, and ranged from 0.10 - 0.15. This
value was then used for determination of all the constants mentioned above.
The results produced by the model were compared against the original catalogue data and Qe comp
was found to match to within ±9%, and Wcomp to within ±6%. Judging these errors against those
reported by other authors provide an indication to the acceptability of the approach detailed. As
an example, Popovic and Shapiro (1995) report errors of greater than ±5% of the measured
power and ±10%, and in some cases greater than -20%, for the capacity.
3.3 Simulation of condenser and evaporator performance

A general model was developed to simulate the performance of a horizontal evaporator and
condenser, of parallel- and counter-flow7geometries. The construction of the exchangers was
based on a multi-circuit concentric tube design, using a liquid HTF. The objective o f this exercise
was to generate fictitious data that is principally identical to what would be expected as if
measured from a standard calorimeter test.

Inputs relating to the physical construction of the exchanger are: number of refrigerant/water
circuits, individual circuit tube length, refrigerant tube internal diameter, HTF tube external
diameter, and ratio of outside (HTF) to inside (refrigerant) surface area. Inputs for the refrigerant
conditions vary between condenser and evaporator. The following pressures and temperature
w7ere fixed for the condenser: dew-point temperature corresponding to inlet pressure, superheat
above saturation temperature corresponding to inlet pressure, subcooling temperature drop, HTF
inlet temperature, and HTF mass flow7rate. Correspondingly for the evaporator, dew-point
temperature corresponding to outlet saturation pressure, TEV inlet temperature and pressure,
superheat temperature rise, HTF inlet temperature, and HTF mass flow rate were fixed. Lastly,
the model allowed ATG characteristics to be specified, as detailed in Section 2.4.3.
It is noted that it was deemed not necessary to validate the models against experimental data
because the trends in performance behaviour are the desired outputs. Output data o f interest
were: incremental refrigerant and HTF temperatures, total condenser or evaporator capacity, and
refrigerant saturation temperatures corresponding to condenser inlet pressure or evaporator outlet
171
pressure. These input and output values were subsequently used to assess the effect of zeotropic
refrigerants on heat exchanger characterisation.
Numerical approach
Based on a set of specified temperatures, calculation of condenser and evaporator capacity' was
based on local heat transfer for both single- and two-phase regions. For the single-phase regions,
the size of each element was based on the enthalpy change (equation 112).
SH = ( 112)
nh
where the term H {out) - H is the enthalpy difference over the entire single-phase region, and
N H is the chosen number of elements. Within the two-phase region, the elements were based on
vapour fraction (i.e., linked directly to enthalpy change) (equation 113).
(H 3)
where the term x (out) - x(in) is the vapour fraction over the entire two-phase region, and N x is
the chosen number of elements. Within the condenser, equation (113) is effectively the reciprocal
1/ N x . For the evaporator, equation (113) corresponds to (1 - xe(iw)) / N x , and inlet vapour
quality was found in the normal way (equation 114).
*>('«)
X e fin) = IT
1-----
---- Tj v 14)'
(1
“ U “ 4 ,j
where the enthalpies H l s and H 4 s correspond to saturated temperature at local evaporator inlet
pressure, and / / 3 at the TEV inlet conditions.
Refrigerant properties were taken from the Refprop database whilst properties of HTF were
obtained from the equations in Appendix 1. Extraction of properties from Refprop was via a link
to its relevant subroutines and the equations in Appendix 1 were included in the programming
code. To acquire different glide size and shape normally require reformulation of refrigerant
composition, but this also results in modifying other thermodynamic and transport properties. To
avoid confusion in this respect, an approach that used a full fictitious ATG with reference to the
dew-point temperature was adopted, where size (equations 103 and 104) and shape (equation
105) of the glide is dictated by the inputs, so that the effects of glide can be isolated and studied.
172
The local HTF temperature ( THTF l ) is obtained from the energy balance (equation 115).
* o,
T ff T F j ~ T H TF (in) ± Z - i - r
(115)
;=1 HTF ' *- P h TF
where N t is the number of elements preceding element / , QJ is the capacity of each of those
elements, and the sign depends upon whether it is heating (condenser) or cooling (evaporator).
For either heat exchanger, the total capacity ( Otot) comprises the heat transfer associated with
the refrigerants’ two-phase region ( Otp), and the single phase regions, which may include liquid
( Qspj ) and/or vapour (Q ) (equation 116).
Q to t - Q s p J + Q s p .v + Q tp (116)
Calculation of heat transfer betw een fluid streams of each element was based on equation (86) for
the NTU-8 method.
min,i (117)
where AT(m), refers to the temperature difference between the two fluids within each element, s
was calculated as equations (98) or (99) according to its geometry, and C min was determined
from the HTF or refrigerant stream (Section 2.4). The total capacity of the single-phase and two-
phase regions of a heat exchanger was based on the sum of its elements (equation 118 and 119,
respectively).
NH
Q s p ^ Q v ,, (118)
1=1
Nx
(119)
To clarify, CHTF i was obtained as equation (88) but adapted to account for the number of
circuits (equation 120), and C'r i from (97), again adapted for the number of circuits and local
change in refrigerant temperature (equation 121).
(120)
173
where N crt is the number of refrigerant circuits.
Here, it is important to note that since pressure losses along the refrigerant flow path were
accounted for, there will always be a difference in refrigerant inlet and outlet temperature even
with a single-component refrigerant. Therefore calculation o f C'r i from equation (121) is
applicable for all refrigerants.13 Since s is a function of NTU, calculation of local U was
required. Neglecting thermal resistance of construction materials U was obtained with the
conventional equation (122).
i i
U; 1--------------- ( 122)
R s u r f '^ H T F ,i )
where Rsurf is the ratio of external to internal surface area, hHTF is the HTF-side heat transfer
coefficient, and hr is the local refrigerant-side heat transfer coefficient, as determined from the
relevant Section 3.3.2 or 3.3.3.
Ultimately, the solution of the heat exchanger model is implicit and convergence of Otot was
achieved through successive estimates of refrigerant mass flow based on determining total
exchanger area. Thus, mr was revised in proportion to the over- or under-estimation o f total
exchanger area, until calculated area matched the total design area (A totdes) (equation 123).
( NH+Nx V /3
mr = mr (0) A [ot,des 1 ^ , A j (0 ) (123)
where mr (0) and j4(0) refer to the mass flow and total area of the prior iteration, and the index
'/3 was used to provide stability to iterative process. Convergence o f calculated area to within
0.01% of Atot des was usually achieved after some 50 iterations. Since mr is unknown, an initial
guessed value was required. The first m r (0) was estimated by evaluating equation (86) in an
13 The exception to this is a unique case in the evaporator only, where the temperature gradient
corresponding to the pressure loss along the refrigerant flow path is the negative of the
temperature glide gradient. Thus, each incremental decrease in temperature due to reducing
pressure is equal to each incremental rise in temperature due to glide, thereby cancelling each
other out causing isothermal flow.
174
energy balance for the entire heat exchanger, whilst assuming Cmin = CHTF, C ' = oo, and an
average U = 100 W m'2 K 1 (equation 124).
f \
^ ( w ) ' CH TF "100 -A totdes~
1 - exp
^ dp,s H b p ,s r HTF
\ J
where H dp s - H bps is the enthalpy difference between saturation points at constant pressure.
The basic calculation sequence was as follows:
- With mr (0), determine local C mm (equations 120 and 121)
- Calculate effectiveness of each element with equation (125) based on capacity from local
enthalpy balance (equation 118).
- From € and Cmw, calculate NTU; for parallel flow (equations 126) or for counter-flow
(equation 127) (transposed from equations 98 and 99).

- Estimate surface area of element (equation 128) and for pressure loss calculation, the tube
length of each S x or S H element (equation 129).
- Sum the calculated area for each element, including single- and tw o-phase regions and revise
mr (equation 123).
a
(125)
C m in,i ' ^ ( i n ) , ; i
ln[l~£, (1+ /?£, )]

N w ,i = --------- — ----- — (126)
[ + K C.i
- 1
Nw , ------ !----- In (127)
% -l v '
■R c>
a, - 1
’ y
N rr; C
A ,= (128)
- (129)
n •D
This procedure is elaborated on in Figure 52 for the condenser and in Figure 53 for the
evaporator.
175
3.3.2 Condenser heat transfer and pressure loss
Throughout the literature there exist a range of different correlations for condensing heat transfer
coefficient. Most were developed specifically for condensation of pure fluids, although some have
been shown to provide relatively accurate results in the case of mixtures. Shao and Granrvd
(2000) provide a correlation specifically for mixture refrigerants, which directly takes account of
the ATg ; this suits the approach adopted for this model where fictitious glide characteristics are
specified.
/'saturated condensing
/ inlet temp. (7).
desuperheating
(A T&,), subcooling
input values
(AT.,). condenser
geometry. HTF details j
± {Thtftmy Mktf)- no. /
calc pc,„, from 7).^,, elements (.Vi)
T
calc Qc^ . (Qi-h, Q<p, and Q:c)
T
calc each element for
desuperheating, two-phase and
subcooling region
T
calc Q T H T F f o u t u ,
T'noutl:
calc refrigerant properties and

—
f call Refprop
temperature at Trl„i., and p^„)j
T.
Calc heat transfer and
calc area (.4,) and tube length (i,)
pressure loss: hrj, hHTFj, U„
for Qh and corresponding
A Cme,j, e„ N-rujy A, and
pressure drop (Ap,)
z„ Ap„ and p(aa)j
calc cumulative area,

= A.„, +
does A,a = AtaAe. ?
Figure 52: Calculation sequence fo r condenser capacity
176
Condensing heat transfer coefficient for pure (and azeotropic mixtures) - where ATG = 0 - is
obtained from the conventional equation (130).
N u-kj
K J N T g = o) = (130)
D
where k, is the liquid thermal conductivity and Nu is the Nusselt number. The correlation for
Nusselt number (equation 131) was derived for pure and zeotropic mixtures.
xl/6
J /3 H2
Nu = c, •Prt R ecp (131)
CP r (Ts - T wall)
where Prl is liquid Prandtl nrnnber, the term H 2s - H l s is the saturation enthalpy difference.
Cp, is liquid specific heat and the term Ts - TwaIl is the temperature difference between
saturation and the tube wall surface ( O, / [hrcJ ■A,]).
When the vapour Reynolds number Rev < 24000, the constants c } = 15.9 and c 2 = 0 .1 5 ;
where Rev > 24000 the constants c, = 0.084 and c 2 = 0.67 . The vapour Reynolds number is
according to equation (132).
0.5
G.. D \ ( p P
Rev = (132)
Pi
where D is the tube inside diameter, /q is liquid viscosity and p } and p v are liquid and vapour
densities respectively. The vapour mass flux ( Gv) is calculated from the mass flow of refrigerant
vapour (equation 133).
4
Gv = x - m , ■ (133)
V - 71 D
where the void fraction ( v ) is calculated from a model developed by Smith (equation 134).
( 1/2 A '
-1
u =U + * ( l - x *•(/>// P v H 0-4 • ( ! - * ) (134)
0.4+ 0.6
Pi x + 0.4-(l-x)
Shao and Granryd identified an inequilibrium of components between the liquid and vapour phase
occur which results in resistance in mass transfer, and therefore offered a compensatory
177
correction factor. Heat transfer coefficient for a mixture - when ATG > 0 - is from equation
(135).
1
hrA A T G >0) = h ( A T G =0) (135)
1 + ^ +F2
where F, and F2 represent the effects of the liquid-vapour inequilibrium (equations 136 and 137
respectively).
F] = x -F {x }l
(CpY) (KJATa =0n
(136)
l o ’ JX Vv J
hr^ A T G =0)
(137)
'X hi-i
where Cp' is the apparent specific heat as already detailed in equation (94). and the functions
F{x}j and F{x}2 are as equations (138) and (139).
F{x}, = l . l - x - ( l - e x p [ - 1 0 - ( l - x ) ] ) (138)
F {x}2 = 0.9-(l-x) (139)
hl v and /jm are the heat transfer coefficient between the liquid interface and vapour, and the
liquid interface and liquid, respectively, and are approximated from equations (140) and (141).
k
h,_v = • (o.023 • Re°v 8 •F rv°3 ) (140)
ht_t = ^--(o.023-Re™ •Pr,03 ) (141)
where kv is vapour thermal conductivity, and Reynolds number and Prandtl number are
calculated in the usual way from vapour ( Rev, Prv) or liquid ( Rel , P p ) properties only.
Pressure loss occurs along the flow path, and it is important to account for this since it affects
local saturation temperature, and therefore influences the glide. The pressure loss within each
element was calculated using the correlation developed specifically for condensation pressure loss
of pure and mixture refrigerants, proposed by Shao and Granryd (2000). Normally, the total two-
phase pressure loss is the sum of frictional {f r ) , momentum (M ) and gravitational ( g )
components (equation 142).
178
For horizontal flow, the gravitational component is not applicable so (d p /d z )g = 0. Also, the
contribution from momentum has a positive effect, but in comparison to the frictional component
it is reported to be miniscule, so (dp / dz)M = 0. Equation (143) was used to calculate the
frictional two-phase pressure loss, which is determined from the vapour-only pressure gradient,
adjusted by a two-phase multiplier (<f>tp).
= t. - f ^ (143)
d z ) fr r * \ d z JSrx
where (dp / dz)fr v is the frictional pressure loss of a vapour-only flow' (equation 144).
dp ] _ _ 2 -Av -(x -G )2
(144)
dZJfr,r
where the vapour only friction factor ( Av) is calculated from equation (145).
/ \ -0 2
x-GD
Av = 0.045 (145)
Mr
The tw o-phase multiplier (equation 146) was correlated to the Lockhart-Martinelli parameter by
Shao and Granrvd using data for both pure and zeotropic refrigerants (equation 147).
<f>t =1 + 0.8- X * (146)
0.9 ( \0 .1
1- x 1 Mi
(147)
\r * v J \P l J
3.3.3 Evaporator heat transfer and pressure loss

For the present study, the correlation of Jung and Radermacher (1993) was selected upon the
recommendation of others who review ed the various options within the literature (e.g., Hewitt,
1999). The tw'O-phase heat transfer coefficient of the evaporating mixture ( h r e ) is from equation
(148).14
14 Note that the following equations differ slightly from the citation due to the published version
containing errors.
179
K * = (Fjr ■Fu„)■ h pb + (Fme -Fv )-h , (148)
where FM is dependent upon the Lockhart-Martinelli parameter (equation 149 or 150).
Fm = 4048• X'tt22 ■Bo' u , for X tl < 1 (149)
Fm = 2 - 0 . 1 - X ~°28 • Bo~°33, for 1 < X tt < 5 (150)
Bo is the boiling number as defined in equation (151).
Bo = --------- ^-----------------------------------------------------
G - ( H u - H 4s)
where the heat flux is calculated for the element from a previous iteration, q = Otp i / , G is
the mass flux, and the term H ]s - H As is the saturated enthalpy difference. The parameter Fun
is a mixture correction factor (equation 152).
k , = {[i+(*, + * 3) • a + * 4 ) i a + * 5 > r d52)
where b2 ...b 5 are constants, defined in equations (153), (154), (155), (156), and (157).
1.01-7,
62 = ( l - 7 7) l n + Y,I In f y0 + (YV~Y,) 1.5 (153)
101 - Y v InJ
b3 = 0, for Yj >0.01 (154)
=(yv / 7 / )0-1 - l , f o r Yt <0.01 (155)
b4 = 152-(ps / p crnn,c ) (156)
b5 = 0 .92-(Yv - Y , r m <P s l Pcr,mvcY (157)
where Y} and Yv are the mole fractions of the more volatile component in the liquid and the
vapour respectively. p s is the local saturation pressure and p crmvc is the critical pressure of the
more volatile component (MVC).
The pool boiling component of the heat transfer coefficient ( hpb ) is from equation (158).
180
(158)
(Yl x / h x +Yl 2 / h 2)
The local heat transfer coefficients, \ and h2 are based on the assumption of the individual
components only. This is calculated using the Stephan and Abrahamson correlation (159).
0.475 t.581
/? = 207 • ■{ q bx ] ■Pr? (159)
[ kj Ts ) p j
where bx is from equation (160) and Ts is the local saturation temperature.
\0 5
2 • cr
bx = 0. 01 46 /? (160)
g ' (P l ~ Pv)
where cr is refrigerant surface tension, and the contact angle, /? =35°. The remaining terms in
equation (148) are found from equation (161), (162), and the liquid-only heat transfer coefficient
( h j ) is equation (163).
= 1 - 0 . 3 5 - ( 7 ,,- ty )1 (161)
Fw = 2.37-(0.29+ 1 /X K) 0.85 (162)
G - ( \ - x ) D \ 01
h, = 0.023- -Pr? (163)
D Pi
Although the model for evaporating heat transfer coefficient (equations 148 to 163) is for
refrigerant mixtures, it may also be used for calculation of pure refrigerants. This is achieved by
setting the composition of the MVC as Y = 1, and in doing so, the mixture correction factors
reduce to unity, i.e., Fun = 1 and Fme = 1. The acceptability of this approach was confirmed in
Jung and Radermacher (1993).
With reference to the method of adopting fictitious ATG within the heat exchanger models, it can
be seen that this approach is suitably handled by the chosen correlation for condensation hr
where refrigerant apparent specific heat is employed (see equations 136 and 137). In contrast, the
chosen correlation for evaporation hr relies on mixture compositions instead of glide
characteristics. This was found to be the case for other such correlations (e.g., ASHRAE, 1997,
Jung and Radermacher, 1989, Gaibel et al, 1994). To maintain consistency, the approach of
181
fictitious ATg was integrated into the evaporation correlation using a method devised as part of
this study. This method was based on linking fictitious compositions of the mixture MVC to the
selected glide characteristics, and is detailed below.
start / saturated evaporating

outlet temp. (Te dp(out>),
superheating (ATsh),
input values evaporator geom etry,
H TF details
mHTF), no. elements
calc Pefoiî from TiAp(oux) (Mr)
guess m.
Calc Qexvapi (Qtfb 2nd Qsh)
calc each elem ent fo r tw o-phase

and supeiheating region
calc Q„ Th
calc refrigerant properties and

tem perature at Trf.„u and
calc heat transfer and pressure

calc area (A,) and tube length (r,)
loss: hrj, hHm„ U„ AT(mU,
for Q„ and corresponding
Cm„,„ s,, N tuj, A, and Ap„
pressure drop (Ap,)
and Pfcuiji
calc cum ulative area.
does i = Ni ?
calc Qe.t
end
Figure 53: Calculation sequence fo r evaporator capacity
From equations (152) to (161), it is seen the two mixture correction factors ( Fun and Fme) are a
function of the mole fractions of the MVC within the mixture. Examination of the relationship
182
betw een MVC mole fractions and glide characteristics - ATG and shape, nG - of several binary
mixtures, indicated that properties could be indirectly linked to the calculation of hr e .
Subsequently, for a chosen glide characteristic the composition of the MVC in liquid (Y{) and
vapour (7V) phases over the range x = 0 —►1 can be estimated.
First, the mole fraction in liquid phase at x = 0 is chosen, which is also equal to the vapour phase
composition at x = 1. As mentioned in Section 2.4, w hen nG = 1, the composition of a binary
mixture was observed to be approximately equal for both components, so Yj(x = 0) - 0 . 5
(assuming similar molar mass) based on analysis data for a number of refrigerant combinations.
A larger nG indicates a low?er proportion of MVC, whereas a smaller nG implies mixture is
richer in the MVC.
As the relative size of the vapour space increases, a higher proportion of the MVC migrates to the
vapour phase, so as x —* 1, Y{ reduces. This shift in composition was linked to the size and
shape of glide, and could be approximated with a power-function (equation 164).
Yl (x = l) = Yl (x = 0 ) - a r • — (164)
n,
where aY = 0.07 and tin = 0.53 based on the refrigerants listed in Table 31 of Appendix 2, and
a greater difference in the components' partial pressures leads to a smaller aY and nn . For a
correct solution of equation (164), aY and nYl may require adjustment to ensure that Yt (x = 1) >
0, always.
The corresponding vapour phase compositions are also needed. Since Yv( x - 1) = Yt (x = 0), it is
necessary to estimate the equilibrium vapour composition at x = 0, and it was found that the
variation in Yv over x = 0 —►1 provided a close correlation with 7Z over x = 0 —►1 (equation
165).
r v( * = o)= j; (x= i)+[ r, ( * = o) - r, ( * = i) ]•«,GnTv (165)
where the index was found to be about nYv = 1.4 for most of the assessed refrigerants.
183
Finally, the liquid and vapour compositions at local vapour qualities were estimated using the
same approach as for local saturation temperature (Tx , equation 104), so equation (165) was
modified accordingly for liquid (equation 166) and vapour (equation 167) phases.
y , W = y , ( x = o ) - [ r , ( x = o ) - y , ( x = i ) ] ( i - [ i - x ] nG) ( 166)
7v(x) = r j x = 0) - [Y,.(x = 0) - Yv(x = 1) ]■ (l - [1 - x]"° ) (167)
These formulas were subsequently used in the calculation of evaporator heat transfer coefficient.
The pressure loss for the mixture refrigerant within the two-phase regions has been accounted for
using the correlation of Jung and Radermacher (1989, 1993) as recommended by others (e.g.,
Wang and Chato, 1995), equation (168).
A
± = ^ r O l . ± . ( ( l . dx (168)
dz D'P j Ax (Jxi
where is the liquid-only friction factor (equation 169), and <f)tp is a two-phase multiplier
(equation 170), which is defined differently from condensation (i.e., equation 146).
Aj =0.046-R e;02 (169)
^tp ~ 12.82 • X ttXA1 ■(1 - x)18 (170)
Local void fraction is required for various heat transfer and pressure drop calculations. ASHRAE
(1997) and Chen and Kruse (1995) give a selection o f different models. Chen and Kruse
compared results from computations against measurements with R23/R152a and found the
Hughmark correlation was most suited, so it w'as selected presently. Equation (171) is the basic
formula for void fraction ( v ).
1
v = a- (171)
( * / A,) + ( l - x ) /
where a is obtained from equation (172).
a = 0.7266-3.482 x 104 • &- 5 ^ 5 . + 0.0601 • 73 (172)

&
where the coefficient 3 is from equation (173).
184
(173)
(1 - 3 / a ) V4
Reynolds and Froude number ( F r ) are from equations (174) and (175), respectively.
(174)
1 ( Gx ]
Fr = (175)
g-G {(v ! a ) - p vJ
Where the mass flux ( G ) is calculated in the usual way7(equation 176).
G=± ^ (176
z-D 2
Evidently, determination of v is implicit and therefore required the calculation to iterate through
equations (171) to (176) until input and output values of v converged. The calculation was
initiated by setting v - x .
3.3.4 Single-phase heat transfer coefficient

Single-phase heat transfer coefficient was required for superheat, desuperheat, and subcooling
regions, or the HTF. It is assumed that fully developed laminar or turbulent flow exists. For a
laminar region (Re < 2300) wiiere there is a uniform surface heat flux (as it assumed in the
model), the Nusselt number ( N u ) is according to equation (177) (Incroprea and De Witt, 1990).
Nu = 4.36 (177)
For turbulent flow (Re > 2300), the accepted Dittus-Boelter correlation (Incroprea and De Witt,
1990) for Nusselt number is employed (equation 178).
Nu = 0.023- Re0* •P rn (178)
where the value of n depends on the whether the fluid is heating ( n = 0.4) or cooling ( n = 0.3).
Reynolds and Prandtl numbers are found in the usual way (equation 179 and 180).
(179)
(180)
k
185
Bulk velocity ( u ) in equation (179) is determined for in-tube flow (equation 181) or from the
flow area between the two tubes for a concentric tube (equation 182) in the case of using a HTF.
m
u —— (181)
p it D~
U= H 1----- (182)
P it-(D ; - D ~ )
Heat transfer coefficient is from equation (183).
(183)
D
Although pressure loss of the HTF is irrelevant to the purpose of the model, it is important to the
single-phase regions of refrigerant flow. The conventional approach detailed in Incroprea and De
Witt (1990) was used.
Jd L ^ P - < (184)
dz 2 -D
where the friction factor is determined by Reynolds number (equation 185).
X =aIR en (185)
where for laminar flow' a = 64 and n = 1, for Re < 20000, a =0.316 and n = 0.25, and for
Re > 20000, a = 0.184 and n = 0.2; throughout, the pipe surface was assumed to be smooth.
Table 25: Heat exchanger simulation input values
Inputs Condenser Evaporator

Refrigerant dew-point temperature +20 to +60°C (40°C) +10 to -40°C (0°C)
Superheat 0 to 60 K (0 K) 0 to 10 K (0 K)
TEV inlet temperature 0 to 5 K below' Tbp (0 K) 20 to 60°C (30°C)
Temperature glide, ATG(p atm) Oto 10 K (10 K) 0 to 10 K (10 K)
Glide gradient; o,, a2 inequation

as equation (103) (0, 0) as equation (103) (0, 0)
(103)
Glide shape, 0.25 < nG < 4 (n G = 1) 0.25 < nG < 4 (n G = 1)
HTF mass flow rate 0.4 kg s"1 0.3 kg s'1
Exchanger length 3m 3m
Number of refrigerant passes 1 to 10 (6) 1 to 10 (6)
Geometry counter, parallel counter, parallel
186
3.4 Observations from component performance simulation
The condenser and evaporator models were used to examine the effects associated with zeotropic
refrigerants on their performance, which were highlighted in Section 2.5. The following presents
detailed discussions of the component behaviour based on the simulation of the heat exchangers,
generated to represent the data expected from calorimeter testing. The model inputs were based
on typical commercial heat exchangers, and are listed in Table 25, and when a given variable is
not being examined it was fixed according to the value in parentheses. In terms of fluid choice,
water was selected as the HTF, and refrigerant properties (other than glide) were based on a
blend of 50% R290 and 50% R600a by mass.
Influence of approach temperatures and variable heat transfer coefficient

The discussion in Section 2.4.2 considered the behaviour of fluid temperature profiles to
understand as the refrigerant dew-point temperature gets gradually closer to the HTF temperature
and to explain the existence of zero virtual capacity whilst a temperature difference is present due
to the glide. Since the condenser exhibits full phase-change of the refrigerant (as opposed to the
evaporator) this will be used in the first assessment.
50
refrigerant
r (m )
HTF
40 - 40
O r(out)
-cs 30
2
8.
o
H
20 -
20
HTF (in)
HTF (in)
10
0.0 0.5 1.0 0.0 0.5 1.0
Distance along exchanger (m) Distance along exchanger (m)
Figure 54: Counter-flow condenser, C \ > C Htf Figure 55: Counter-flow condenser, C ’r ~ C htf
Figure 54 show s temperature profiles for a large 5Tc(m) of 30 K. Since C'r > CHTF the HTF
temperature change is larger and the temperature profiles taper together as HTF approaches the
refrigerant inlet. When ATc{in) was reduced to 20 K, a point is reached where C ' ~ CHTF,
resulting in an equidistant temperature profile throughout the exchanger length (Figure 55).
187
Figure 56 illustrates the temperature profile where ATc^ was reduced to 10.1 K, resulting in a
situation where C ' « C HTF . Observation of the data-points corresponding to equally
incremented changes in vapour quality (i .e. spacing of the data points) shows a pronounced shift
of heat transfer towards the refrigerant inlet. Since the temperature difference is minimal towards
the refrigerant outlet, the majority' of ATG is ‘‘pushed” towards the refrigerant inlet end of the
heat exchanger with the greatest AT with majority of heat transfer occurring within one-third of
the exchanger. Extrapolating this behaviour until ATc(j n^ is reduced to its minimum wfren Tc{out)
—> TffTF(in) , all heat transfer is pushed to towards the refrigerant inlet (Figure 56) and eventually
no heat transfer can occur when ATc^ = AT- . The entire ATq will be forced to the point at
which the refrigerant enters of the exchanger. Simultaneously, the refrigerant mass flow
necessary to give up an infinitesimally small quantity of heat converges on zero mr .
50 21 T
refrigerant
HTF
40
oG O
0 <oul)
¥5 ¥
r(m ) 1 20
2 30 HTF (out) HTF (in)
6p
r(oul)
20
H TF{out)
H TF(in)
19
10
0.0 0.5 1.0
0.0 0.5 1.0
Distance along exchanger (m)
Figure 56: Counter-flow condenser with very Figure 57: Counter-flow condenser with pure
small approach, C \ « C htf refrigerant and a small A T C(in)
Parallel-flow condensers were found to behave with identical manner, where the refrigerant enters
at the same point as the HTF where the initial temperature difference is greatest produces the
majority of heat transferred. As both fluids progress along the exchanger length, the temperature
difference reduces until they meet at the exchanger exit. This behaviour similarly applies to the
evaporator; a ATe{m) —►0 K, mr and Qe reduce and the majority of the heat transfer occurs at
the refrigerant inlet where the temperature difference is greatest. Eventually when Te dp —>
20
|
.•s 10 -
eo*
o8-
ATCr = 10K
60
Dew-point temperature (°C)
Figure 58: Characteristic line fo r condenser with pure and zeotropic refrigerant
For comparison, exchanger temperature profiles using a pure refrigerant are shown in Figure 57,
where ATc{mj was gradually reduced until a minimal temperature difference was achieved as in
Figure 56. Obviously, the refrigerant does not exhibit a change in temperature, and as a result, a
near-even distribution of the equally incremental x data-points is seen which indicates that a
similar amount of heat is transferred across the total exchanger length, even as O —►0.
32
30
28 -U
26 nn =1/3
24
22
20
HTF
18
0 0.2 0.4 0.6 0.8 1
Figure 59: Influence o f glide shape on temperature profiles fo r small ATc(in)
The simulation was also used to illustrate the condenser characteristics using a pure refrigerant
and a zeotrope (Figure 58). An initial observation is that neither characteristics are linear, which
equates to a non-constant basic rating. However at high ATc(in) a pure refrigerant gives an
189
approximately constant basic rating. For the zeotropic refrigerant, the curvature tends to span
over a wider range, giving a variable basic rating.
Effect of non-linear glide shape on tem perature profiles

The influence of the glide shape on temperature profile and exchanger characteristic was
addressed. Figure 59 illustrates how the shape of the glide affects the temperature profiles. It can
be seen that as the shape moves from concave ( nG = 3 ) to linear to convex ( nG = !A), the
distribution of the x elements are pushed towards the higher AT at the refrigerant inlet. Where
nG = Vs, a significant proportion of ATG and heat transfer occurred within one-tenth of the heat
exchanger. Conversely, when nG = 3, there is a broader distribution of heat transfer.
The variation in glide shape produces a change in capacity despite a fixed ATc(w). Shown in
Figure 60 is the relative capacity' of the condenser for a range o f shapes over a series of ATc(m).
For individual glide shapes, the influence on capacity' is greatest at small ATc(m). For convex
glide shapes the capacities are always larger than that of the linear case, whilst concave shape
always produces a lower capacity. In the case of the evaporator the opposite is true concerning
the impact of glide shape, i.e. when nG < 1, the capacity is greater than when nG = 1.
3
11
co
©
? 2
^5
£
'1
§• 1
O
= 1/ 2
0
10 12 14 16 18 20
Approach temperature difference (Tcdfl - T HTFim))( K)
Figure 60: Variation o f capacity with ATc(in)fo r different nG
Effect of sensible heat transfer

In reality a proportion of the condenser or evaporator capacity is due to sensible heat transfer,
and this is normally small compared to the two-phase region, and because h is lower a relatively
large exchanger area is required. On this basis, it is useful to assess the condenser and evaporator
performance in light of sensible heat transfer, and is illustrated in Figure 61, which shows
condenser temperature profiles with 30 K desuperheat and 3 K subcooling. The high use of
surface area is confirmed where the sensible heat transfer is seen to use approximately one-third
of the exchanger area (distance), whilst representing a much smaller proportion of the total heat
transferred. A similar effect was found with the evaporator. The implication is that O reduces
disproportionately and thus influences basic rating.
Sensible heat transfer also influences the minimum approach temperature difference. As
demonstrated previously, AT(m) may be gradually reduced until ATminiin) is reached, and this
corresponds to ATa when considering phase-change region only. If a condenser comprises
subcooling ( ATsc) then an additional offset is produced for ATcmin(in) (equation 186). Similarly
with the evaporator, existence of superheat ( ATsh) w ill also offset ATemin{in) (equation 187).
c,min(in) = ATG +ATSC (186)
e,min(in) - ATG.e + ATsh (187)
refrigerant
desuperheat
subcooling
r(out)
HTF (out)
HTF (in) J
10
0.0 0.5 1.0
Distance along erchanger (m)
Figure 61: Condenser temperature profile with single-phase heat transfer
It is noted that whilst the presence of desuperheat ( A ) within the condenser does introduce a
further offset, the definition of AT(in) is based on the inlet saturation temperature - Tcdp - so in
this context it does not directly affect ATcmin(in).

Figure 62 shows various condenser characteristics based on Tc dp. to illustrate the influence of
ATsc or/and A 7 ^ . In terms of approach temperature, it is seen that without subcooling,
A T c,min(m) = A T g , i e - 10 K > but when subcooling is included, ATcrmn(in) = ATa +ATSC, i.e. 13 K.
Regarding the effect on condenser capacity, inclusion of desuperheat or subcooling or both

produces an overall reduction for a given ATc(m). The greatest reduction occurs when both ATSC
and A a r e included. It is also seen that the actual shape of the overall characteristic becomes
more linear.
two-phase only
o Arj(, = 30K
-a- ATj, = 30K, =3K
- o- atk =3K
O 10
20 30 40 50 60
Figure 62: Condenser characteristic affected by single-phase heat transfer
Effect of pressure loss

Pressure loss along the refrigerant flow path produces a reduction in saturation pressure and thus
a change in local saturation temperature. The simulation was used to observe the subsequent
effect of relatively large Ap (for clarity) on local saturation temperature ( Tr ), Figure 63. The
dashed lines represent T+ and that correspond to the inlet pressure. Ap along the flow path
causes these saturation temperatures to fall, Tc dp (local) and Tc bp (local), and the corresponding
local Tr is shown passing from the inlet Tc dp to the outlet Tc bp. Whilst Tr lowers as a result of
ATg , an additional reduction occurs simultaneously as pressure falls (in this example, equivalent
to about 5 K in Tc dp). The majority of the contribution from Ap is seen to occur in the first half
of the condenser, where the high vapour quality produces the greater effect. The effect is opposite
in the evaporator; whilst Ap reduces local saturation pressure ATG is partially neutralised,
rather that complimenting the glide. For both condenser and evaporator, the effective shift in
192
ATg alters C ' and therefore B , thus this phenomenon should be accounted for within the
revised exchanger characteristic.
50
r(in)
40
30
20
0.0 0.5 1.0
Figure 63: Effect o f pressure drop on refrigerant temperature in condenser
Figure 64 presents a condenser characteristic (for two-phase only), where the reference state
corresponds to Tdp at condenser inlet pressure ( Tcdp{m)) and also Tdp at outlet pressure
(Tcdp(out)). With ATc(in) increasing, Oc, mr and thus Ap increases which is manifest in a
greater differential between Tc dp(m) and Tcdp(out). The existence of two possible characteristics
indicates the importance of defining the appropriate reference for rating purposes. Because
the condenser is ultimately to be linked to the compressor discharge, then the correct location for
the reference has to be the condenser inlet, i.e., Tc dp(in). Similarly, the evaporator has to be
linked to the compressor suction and therefore the correct location for the evaporator must be
the evaporator outlet, Te dp{out).
A further complication is introduced when considering the effect of Ap within the sensible heat
region. In the case of desuperheat, Tdp at the onset o f condensation will be lower than that
implied by the Tdp that corresponds to the pressure o f the superheated vapour at the inlet of the
condenser. Nevertheless, to be consistent with the reference condition at the compressor outlet,
Tc,dp(m) must accommodate the single-phase pressure drop. Similarly, the evaporator
corresponding to outlet pressure will be lower than T* at x = 1 due to Ap in the superheat
193
region. Therefore, Te dp(out) includes the contribution of single-phase pressure drop within the
superheat region.
10 -
30 40 50 60
HTF (m) Dew-point temperature (°C)
Figure 64: Effect o f pressure loss on condenser characteristic (two-phase only)
Aspects unique to the evaporator

Additional consideration is required for the evaporator in terms of the effect of intersection of
ATg due to its dependency- on 77ra.,,(m). The simulation was used to understand the consequences
of this factor.
HTF
-10
u
-20
C
u.S
<L>
C
L-30
£
£
O O O o OOo o ❖
O O o oo oo o0 >refrigerant
-40
-50
0 2 4 6 8 10
Figure 65: Effect o f different inlet temperature on evaporator capacity
Figure 65 illustrates two sets of temperature profile data, where refrigerant enters the TEV at
+20°C and also at +60°C. The former condition produces a lower xe(m) than the latter, and this is
reflected in the size of their respective ATG, although it must be cautioned that the full ATG is
194
obscured by differing contribution from pressure drop. The evaporator ATa is greater when
Tjei'U") = +20°C, thereby producing a larger AT(in) and hence higher Oe which is evidenced by
the lower HTF outlet temperature.
Table 26: Contribution o f TEV inlet temperature to factors affecting capacity
Low TTEV(w^ High Tt

Parameter
Contribution Effect on Oe Contribution Effect on Oe
X e(m ) —0 - 1 -
T e(tn) lower increase higher reduce
Ap smaller increase larger reduce

U smaller reduce larger increase
Larger reduce smaller increase
>
o
Apart from ATG intersection, there are other factors associated with the evaporator inlet
conditions that can cause a variation of Oe . Since hr varies with vapour quality, a higher
TTEV(in) changes U and therefore Oe . Further, as x e(in) —» 1, higher Ap occurs and therefore
amplifies the corresponding reduction in local Tr . Contrary to the condenser, this manifests as a
smaller ATGe across the evaporator. A summary of these effects is detailed in Table 26;
increasing TTEV(m) lessens ATG e due to both effects of intersection and higher A p , whilst
average U may be greater or smaller which would manifest as a variation of Oe .
6 TEV {in) -6 0 °C
5 a ^TEV (in) ~ 3 0 CC
£
4
5o
3
<s
(J
2
1
0
-40 -30 -20 -10 0
Figure 66: Effect o f different inlet temperature on evaporator capacity
195
An example of the net effect on Oe due to the different contributions from TTEV(in) can be
observed in Figure 66. With smaller ATelm), a TTEV{in) = 60°C produces a Oe less than when
TTEV{in) = 30°C, although at larger ATe(in) the case is reversed as the greater Oe occurs when
TTEViw) - 60°C. This cross-over represents the point at which the positive influence of lower
7eiin) ar,d A/? dominate over the negative influence of lower U and larger ATG e .
3.5 Proposed characterisation methods

The purpose of this section is to offer possible alternative rating methods for when components
use zeotropic refrigerants. Existing methods mainly use linear line-fits to describe the
performance of evaporators and condensers, and polynomial equations for the compressor.
Characterisation methods can go further than this by adopting more complex models. Figure 67 is
an overview of the range of possible options available for component characterisation, and the
degree of complexity typically indicates its accuracy. An approach based on functional line-fitting
and approximate correlation with specific variables was chosen as a compromise betw een
simplicity and accuracy. Fictitious test data obtained from the simulation models in Section 3.2
were used to assist in the development of the new methods.
j
linear line-fit sim plistic to
apply, lim ited
range o f
non-linear line-fit (polynomial) application
functional correlation o f perform ance w ith selected variables
sem i-em pirical overall In AT or N T U -e approach com plex

m ethodology,
accurate across
sem i-em pirical com prehensive m odelling w ider range
Figure 67: Range o f performance prediction methods fo r component rating
3.5.1 Compressor
The general expressions for evaporating and condensing capacities as a function of Te dp and
Tc dp are given in equation (188) and (189) respectively.
196
(188)
c,comp
(189)
As discussed, it is also necessary to characterise the degree of discharge superheat ( A7 ^ ), also
as a function of the operating conditions (equation 190).
(190)
Applying these general expressions to the established form (equations 76 and 77) leads to
equation (191) for evaporating capacity, and equation (193) for compressor power. Although the
compressor capacity is normally calculated from Oc comp ~ Q e c 0 mp + Wcomp, a specific formula is
provided here for convenience (equation 192).
e ycomp
(191)
c,comp
comp
where the coefficients c, ... c)0 are specific to Qecomp , Occomp and Wcomp and obtained from
reduction of measured data, Te dp is saturated dew-point temperature corresponding to suction
pressure, and Tc dp is saturated dew-point temperature corresponding to discharge pressure.
Although not currently available from catalogue data, A i s known to vary according to
operating conditions, and the same form was also found to be appropriate in this case (equation
194).
(194)
^8 1 c,dp ^ e,dp 9 ^ e,dp
e,dp ' ^ cc.dp
.d p "*” ^ 1 0 " ^ c,dp
where the coefficients c} ... c10 are obtained from reduction of measured data.
197
3.5.2 Expansion device
The only influence identified was with respect to selection of the appropriate reference conditions
for determining refrigerant enthalpies. Therefore, it was concluded to maintain the existing
characterisation method, although to ensure consistent definitions, i.e., equation (81).
3.5.3 Heat exchangers

Of interest to the performance testing, rating, catalogue presentation and selection stages are the
following:
- ATmjn(in) is greater than zero, since it is offset by ATG and ATsc in the condenser and ATsh
in the evaporator.
- The properties of ATG and the other parameters associated with the exchanger configuration
mean that the condenser and evaporator characteristic is not normally linear, so B is rarely
constant.
- Sensible heat transfer within the condenser affects the capacity which is important because
compressor discharge superheat varies with Tedp and Tcdp, thus the condenser rating should
be a function of A .
- Pressure drop in the heat exchangers require that the reference states must be correctly
chosen to ensure they are consistent with the compressor inlet and outlet conditions; Tc dp{in)
and Te dp{ouf) must be employed in component rating.
- Evaporator capacity is subject to variation according to T1EV(iri) due to intersection of the
ATg , thus TTEV(w) should be accounted for in evaporator rating.
As a basis for the revised rating method, two general expressions for heat exchanger capacity7can
be made. For a fixed geometry condenser, the capacity is a function o f three parameters (equation
195).
= A T ' J ^ A T m , A r„ J (195)
Similarly, the capacity7of a fixed geometry7evaporator can also be expressed as a function of two
operating parameters (equation 196).
Q e,evap ~ f ^ 'e , d p { o u t ') ^ T E l '{ i n ) ) (196)
198
The conventional rating approach for both condenser and evaporator, based on the NTU-s
method, was chosen to develop a revised rating method. A general rating method was first derived
for any heat exchanger using a zeotrope, which was then adapted to account for particular
characteristics of the condenser or evaporator. Through observing the trends in e and Rp , it was
found that ln(f) could be accurately approximated as a linear function of ln(/?c ) over a typical
range of AT{m}, although the gradient when Cmm = CHTF was distinctly different from when
C mm = C ' . This was found to be applicable to the three main exchanger geometries, except for
the unusual situation of —>0 in any exchanger type (i.e., a very small ATG). In response to
this, it was deemed necessary to develop two general rating equations applicable to each case of
C mm ~ C HTF 0 F ^ mm ~ ^ r •
Initially, the original calculation of exchanger capacity were simplified (equation 197) where the
capacity is for the two-phase region only.
Otp=B- AT{w) (197)
B is the product of s and C^ . In the first case, the derivation is for Cmm = CHTP (equation
198).
(198)
The approximate linear relationship between ln(s) and \n(R^ ) as noted above, was substituted
into equation (198), to give equation (199).
B = a - R nd • CHTF (199)
where R^ is the ratio of CHTF to C ', and C ' was shown (Section 2.4) to be equal to Otp / ATG
when based on an entire two-phase heat exchanger (from equation 95). Equation (199) can be
rewritten as equation (200).
(200)
Substituting equation (200) back into equation (197) and rearranging, provides an explicit
expression for the basic rating (equation 201).
199
s n /( n + l)
B = a 1' ■CHTF (201)

AT(ins
Correspondingly, by substituting equation (201) into the initial rating equation (197), a final
formula for capacity is found (equation 202).
CL = a ]/{n+]) •CHTF • A J^/(”+,) • A7;!/(”+

(in) 1)
(202 )
For the second case when Cmm = C ', swapping C ' for CHTF in equation (198) gives equation
(203).
B =e C r (203)
Using the same power-law expression for s and as above, and substituting into equation
(203), gives equation (204).
B = a - R nd -C'r (204)
Now that = C ' / Cm F , and again using C ' = Qtp / ATG for the entire two-phase heat
exchanger, equation (204) was rewritten as equation (205).
Y a
B =a (205)
V ^ H T F ' ■A^G )
AT-
Substituting equation (205) back into equation (197) and rearranging, provides an explicit
expression for the basic rating (equation 206).
(n + l)/n
1 A
B= .1 I n
c HTF (206)
AT(in,
v ("») J
Finally, by substituting equation (206) into the initial rating equation (197), a final formula for
capacity is obtained for when Cmin = C ' (equation 207).
O = n ~ l!n - C ■A 7 ’(n+1)/" . AT71' ” (207)

\ltp u ^H T F a jG a 1 (in)
When using equations (202) and (207) for characterisation of a condenser or an evaporator, it is
useful to know the minimum number of performance data needed to determine the values of a
and n . Assuming that over the intended operating range Cmm = CHTF always, then a minimum
200
of two O - ATm measurements would be needed, and the same applies if Cmw = C ' over the
intended range of operating conditions.
A small ATG implies a larger C'r , which suggests that generally C mm = CHTF for most small
glide refrigerants. If C'r ~ CHTF over the application range, three O - AT(in) measurements
instead of four would be required. This is because the value o f a in equations (202) and (207) is
the same; the form c = a ■R" implies the values of e for both Cmin = CHTF and C mm= C '
converge when R^. = 1. Nevertheless, since the power-law formula is an approximation, it would
be pertinent to determine a and n from a larger number of data-points.
Condenser
If equations (202) and (207) are to be employed for a rating method, the}- should be modified to
account for the sensible heat transfer regions and pressure drop. First, the generic rating
equations were adapted to account for the contribution of subcooling. Normally, ATsc is assumed
to remain constant over the entire operating range, and given that the liquid refrigerant Cp is
usually constant over a typical temperature range implies is proportional to the refrigerant
mass flow. Assuming that the refrigerant exhibits a proportional change in saturated enthalpy
difference over the expected operating Tc dp, the contribution o f Qsc can be approximated from
equation (208).
Qsc =bsc’Qtp (208)
where bsc represents the basic rating of the subcooled region for a fixed ATsc. Although it was
stated that subcooling is assumed to be constant, equation (208) could be easily modified to
account for variations by changing bsc proportionally, since bx / ATsc ~ constant.
Similarly, the capacity of the desuperheat region was handled independently. Unlike subcooling,
ATdsh is subject to variation because of compressor discharge conditions, and therefore is
not solely dependent upon refrigerant mass flow. Further, as A7 ^ varies the surface area
available for two-phase heat transfer is affected, and therefore the overall condenser capacity' for
the chosen A7^;>l) will differ. It was found that for either a fixed Cmjn or a fixed A t h e r e was
201
a close correlation between the respective f{ A T c{m)) term in equation (202) or (207) and
(equation 209).
(209)
where was found to vary proportionally with A , and was roughly constant (for
each Cmin case). By including a linear variation of capacity with A , and putting AT"(in) =
A7i S +1) for when Crmn = ^ hTF > putting A T ”(m) = A T^” for when Cnnn = C ', a general
rating formula was produced for desuperheat capacity7(equation 210), which was adopted to
enable the integration of desuperheat into the condenser rating.
(210)
where (obtained from a ^ h- AT^ ) and are specific to each Cmm case.
The final parameter for inclusion in the condenser rating is pressure drop. Given that the general
rating formulas (equation 202 and 207) are dependent upon ATG, it is convenient to adjust ATG
accordingly. Thus, the effective temperature glide ( ATq ), which is the difference between
isobaric ATG and equivalent change in saturation temperature ( ATc AP) (equation 211), can be
applied to equations (202) and (207).
(211)
A basic formula for estimating ATc AP was deduced from relating Ap with capacity, and the
subsequent shift along the saturation curve. From Bernoulli, Apecu^ assuming refrigerant
properties do not change significantly over the anticipated operating range. Similarly, for a fixed
geometry velocity can be substituted for mass flow', so Apccm2r . Ignoring any variation in
enthalpy difference, the pressure drop can therefore be approximated as a function of Oc

(equation 212).
Apc =c, Qc (212)
Over a limited range of operating conditions, the equivalent change in Tc dp produced by Apc can
be estimated since ATdp I Ap ~ constant, as w7as observed for most refrigerants. Therefore,
applying this to equation (212) relates change in saturation temperature to capacity (equation
213).
202
^ c ,d p ~ C\ ' C 2 ’ Q c (213)
Using equation (213) within the general rating formulas would result in an implicit equation, so
instead it was chosen to adopt a non-functional line-fit to be substituted for Oc . Taking the
approach of the conventional basic rating method, and a simplified version of equation (210),
provides an estimation of Qc (equation 214).
Qc =ATc m -(bl + b2 -ATdsh) (214)
It is noted that the contribution of the subcooling region is neglected, since liquid flow is known
to produce a minimal Ap relative to vapour or two-phase flows. Lastly, substituting equation
(214) into (213) provides a means of predicting the corresponding change in Tc dp as a function
of ATdsh and ATc(m} (equation 215).
A 7 U =bc^ - ( A T c m •[*, +b2 ■A r « ] ) 2 (215)
where bcAp = c, ><c2 .
In addition to capacity rating, the dew -point temperature corresponding to the outlet pressure of
the condenser must also be provided to meet the requirements for system balancing. The
equivalent pressure loss is expressed as a temperature difference ( ATc AP) (equation 216).
2 c,dp(out) ~ rf'c,dp(in) ^ ^ c ,A P (216)
The final rating equation is obtained by substituting the various adjusted parameters from
equation (211) ( ATG —> AT ’G) and equation (208) for subcooling, and adding equation (210) for
desuperheat, in equations (202) and (207). This provides general rating formulas for: (1) when
Cmm = Cjjjp (equation 217), and (2) when Cmm = C ' (equation 218).
=V. + s;,- ( A T i y '^ - A T ^ (217)
2=W -(Ar ^ ) M -2 + B ’c2_ • Arj,'"2 (218)
where the desuperheat basic ratings and are found from rearranging equation (210)
for each C min case, and the basic ratings B'c ] and B'c 2 are essentially constants, calculated from
equation (219) and (220) respectively.
203
ri/(«i+i) (219)
K i = ( 1 + ^c,i)-tf
(220)
For clarity the temperature difference, ATc{jn) = Tcdp(in) - THTF{m).
The use of degree of desuperheat as a variable also implies that more data is required. In Section
3.5.3, it was noted that a minimum of two or three Oc - A7Vn) measurements would be needed,
depending upon the variation of Cmin. However, since A i s a dependent variable, at least one
additional measurement would be necessary' for each ATc{in) in order to determine . The
constants associated with pressure drop calculations demand that the pressure is known for both
inlet and outlet conditions. Lastly, constants associated with subcooling can be found from
enthalpy balances.
Evaporator
As with the condenser, equations (202) and (207) are to be modified to account for the influences
of evaporator inlet condition and ATG (equation 196) as well as pressure drop. Additionally,
because equations (202) and (207) are based on the two-phase region only it is again pertinent to
adapt the rating method further to integrate the sensible heat transfer from the superheat region.
Consistent with the subcooling region of the condenser, superheat capacity ( Osh) was found to
relate directly to the two-phase capacity (equation 221).
Q s h ~ b s h ' Q tp (221)
where bsh represents the basic rating of the superheat region for a fixed ATsh. Whilst the use of a
TEV or similar expansion device implies a fixed superheat, equation (221) could be easily
modified to account for variations, since bsh / ATsh ~ constant.
Use of equations (202) and (207) allow the influence of variable ATa e to be integrated into the
rating method. Therefore, ATGe must be determined and two approaches are proposed based on
the knowledge of TTEV(w) and Tedp, and the glide properties specific to the refrigerant in use. The
most straight-forward approach is to employ a non-functional line-fit, and for all of the mixtures
204
evaluated good accuracy was obtained with a linear relationship when it contained a smaller
proportion of MVC (equation 222).
AT3 e —ax + a2 •TTEV(j„) + Te dp(jn) • + a4 •TTEJ,(jn^) (222)
where ax to a4 are determined from a refrigerants' property data, and Te dp{in) represents the
dew-point at a pressure corresponding to the evaporator inlet.
Conversely, ATG of zeotropes whose composition is dominated by MVC tend to be independent
of Te dp due to the vertical p - T gradient of the bubble-line, so ATG normally exhibits a non
linear relationship with TTEV(tn). If this is the case then a second-order polynomial equation (223)
can be used instead of equation (222).
(223)
where ax to a, are also determined from refrigerant property data.
Alternatively, the approach detailed in Section 2.4.3 can be employed, where ATGe is estimated
from xe(m). Determination of xe(in^ may be done in the usual way from enthalpies (equation 114),
or by using a non-functional linear relationship based on a temperature drop across the expansion
device (equation 224).
'e(m ) ~ a \ + ° 2 '( J T E V ( in ) T 'e dp{irt) ) (224)
where ax and a2 are from refrigerant property data.
Equation (222), (224), and the calculation for ATe(in) require knowledge of the saturated
conditions at evaporator inlet, whilst the rating data should be based on outlet conditions
( ^ e,dP(out) )• Therefore it is important to establish the equivalent saturated temperature change
across the evaporator due to pressure loss ( ATeEp). This was handled in the same way as the
condenser pressure drop. As with equation (211), the effective temperature glide ( AT 'a ) can be
determined from equation (225).
(225)
205
Similarly, the derivation of ATe hp followed the same sequence as for the condenser; equation
(226) instead of (212), and equation (227) instead of (213).
Ap, = c , - g ; (226)
A T . ^ c , -c2 - 0 ) (227)
As with equation (214), total Oe was obtained only as a function of a single temperature
difference, rather than accounting for sensible heat regions separately (equation 228).
Qe ~ RTF (in) Pe,dp(out) )-b, (228)
Lastly, substituting equation (228) into (227) provides a means of predicting the corresponding
change in Te dp (equation 229), and thus the inlet dew-point (equation 230).
ê,Ap = be,Ap ' ( THTF(in) ~ ê,dP)out) ) (229)
where be Ap = c} * c2 * b{ .
^ e,dp(in) ~ ^ e,dp(out) - A7; ap (230)
The final rating equation is obtained by substituting the various adjusted parameters from
equation (225) ( ATG —» A7^) and equation (221) for the superheat region, into equations (202)
and (207). This provides general rating formulas for: (1) when Cmn = CHTF (equation 231), and
(2) when Cmjn = C'r (equation 232).
=K> -&t ^ <23d
Q .,^ p,2 = K ,i ■ )("2tl>'"2 • AT " g (232)
where the basic ratings B'e l and B'e 2 are essentially constants, calculated from equation (233)
and (234) respectively.
K ) = 0 + K i ) •«y(",+1) •CP htf • (233)
Be 2 —(1 + bsh 2 ) •oe •Cpjjjp • bifjTp (234)
For clarity, the temperature difference, ATe(w) = TH7F(in) - (Te dp(out) - ATq ).
206
As with the condenser, determination of the constants and indices for the new rating equations
necessitate additional performance data-points. Therefore, a minimum of tw o or three Qe -
ATe(w) measurements would be needed, depending upon the variation of C mm. Unlike the
condenser though, additional measurements to accommodate sensible heat contributions are

unnecessary since ATsh is assumed constant. The constants associated with pressure drop
calculations demand that both inlet and outlet pressures are known.
207
SECTION 4: SYSTEM BALANCING
4.1 Introduction
It is the aim of this section to identify the differences associated with balancing system
components when moving from a pure refrigerant to one with a ATG and ultimately devise a
revised balancing technique to improve the accuracy of matching system capacity7with the
application requirements. Both pure and mixture refrigerants are in common use and therefore the
balancing technique should be suitable for both.
In addition, other system operating parameters such as heat exchanger pressure loss and variation
of compressor discharge superheat should also be incorporated into the new balancing scheme.
Based on the principles of system balancing for refrigerants with temperature glide, a general
methodology' should be detailed to allow1the utilisation engineer to apply it.
4.2 Balance technique theory

Normally when balancing a refrigerating system, the compressor and condenser are combined to
produce a characteristic for the sub-system usually referred as the condensing unit. The
condensing unit is then combined with the evaporator to form a complete sy stem (given that the
TEV is ignored, as noted previously). Combining these components is done on the basis of
defined refrigerant reference state temperatures and heat transfer capacities. For a specific
component, capacities at corresponding reference temperatures are obtained from component
catalogue data.
For the methodology described in this section, the following are assumed to apply:
- Basic system. Only a simple system - using a single stage of compression and only four
fundamental components (evaporator, condenser, compressor and expansion device) - is
considered.
- Expansion device. The type of expansion device is restricted to TEV.
- Heat and pressure loss. It is assumed that there are no losses to the environment from the
system, except for those accounted for within the component rating, i.e., no heat transfer or
pressure losses associated with interconnecting pipework, etc.
- Liquid subcooling. The subcooling at the condenser outlet is fixed with the same value for
each component. Component ratings must be based on the same subcooling for compressor,
208
condenser and evaporator. To simplify the current exercise, it is assumed that ATsc = 0 K
always.
- Reference state temperature. The rating reference temperature for all components must be
fixed.
- H TF temperatures. The condenser and evaporator HTF inlet temperatures and mass flow
rates are fixed.
For conventional system balancing using a pure refrigerant, the component characterisation are
expressed in functional form: evaporator capacity (equation 235), condenser capacity (equation
236), and compressor evaporating (equation 237) and condensing (equation 238) capacities.
Q ,^ P= f ( L ) (235)
Q ,c« = f{T c) (236)
Q.*o»r = . f { T , J c ) (237)
Qc,comp = f ( T , . T c ) (238)
The condenser and the compressor are combined to form the condensing unit sub-system using
their respective Oc and Tc variables. Its capacity is characterised in the form o f equation (239),
which shows that the change in evaporating temperature is dependent of the change o f condensing
unit condensing capacity (Oc c u ) with Tc . Te is implicitly dependent on the characteristic.
Q c.".(T .) = f { T c ) (239)
The graphical interpretation of this is shown in Figure 68.
^ condensing
unit
Figure 68: Graphical interpretation o f equation (239)
209
Since Oe comp is a function o f the same two variables as Oc comp, the equivalent evaporating
capacity o f the condensing unit ( Oe c u ) can be expressed in the same terms (equation 240). This
represents the variation in Oe c u with Te, whilst Tc is now implicit.
Qe c u (Tc) = f ( T e ) (240)
Finally, the condensing unit (equation 240) is matched to the evaporator (equation 235) to
determine the equilibrium conditions o f the complete system. The balance point for Te and Oe is
found when the expression in equation (241) is satisfied.
Q e c u i T c ) O e
/ v =° (241)
f(r.) f(T.)
After obtaining the balance Te and Oe , the corresponding equilibrium Tc and Oc can also be
determined. This is achieved in the same manner as conversion of the condensing unit
characteristic, where compressor data is used to find the equivalent Tc and then Oc .
4.3 Balance technique for zeotropic refrigerants

In order to apply the balancing theory to components operating on zeotropic refrigerants, the
influence of ATG on individual component characterisation must be known. To start with, the
desuperheat and pressure loss aspects are ignored at this stage (referred to as the interim
balancing technique), so the heat exchanger inlet and outlet conditions are equivalent, i.e., Tdpiln)
= T = T
dp(out) dp ■
Functionally the compressor capacity' is expressed as before in equations (188) and (189). For the
condenser, Oc cond is based on equation (195), and ignoring A a n d ATx it reduces to a
function o f Tc dp only, equation (242).
(242)
The evaporator capacity has been shown to vary with inlet conditions, the evaporator
characteristic in equation (196) is reduced to a function o f Te dp and TTEV(m), equation (243).
210
However, assuming ATx = 0 K , then TrEV(m) in equation (243) can be substituted for the
condensing dew-point temperature, since Tcdp = TTEV{m) + ATG (equation 244).
( 244 )
The avoidance of TTEV{m) (or Te(in] or Tcbp ) simplifies the procedure.
Following the conventional procedure, the condenser and the compressor are initially combined to
form the condensing unit. Condensing unit condensing capacity (O c CiU) is expressed as equation
(245), and as usual, the evaporating temperature ( Te dp) is implicit within the condensing unit
characteristic.
Q ,c .(T,Jp) = f { T cJp) (245)
Because Qe comp is a function of both Tedp and Tc Jp, Oc c u can be converted so that the
evaporating capacity of the condensing unit (Oec u) can also be determined (equation 246). This
process is consistent with the procedures associated with the conventional technique.
Q ., STcJp) = . f{Tl Jp) (246)
Normally, at this stage the evaporator would be balanced with the condensing unit to form the
complete system. However, from equation (244) it is known that Oe evap is dependent upon an
additional variable, being Tc dp. Since Tc dp is implicit in equation (246), it follows that Oe es,ap
cannot be matched directly to Oec u . In order to match the evaporator to the condensing unit, this
additional variable must somehow7be integrated into the evaporator characteristic. The solution
w as achieved through the development o f an additional sub-system. Since Oe comp is a function of
the same variables as Oe ex,ap, the two components can be combined to produce a new sub-system,
termed the "'evaporating unit'’.15 The evaporating unit evaporating capacity ( Qe e u) is expressed
15 Note that the compressor characteristic has been used twice, once to produce a condensing unit and
now for an evaporating unit. It should be understood that the compressor characteristic is simply
employed to fix a relationship betw een capacities of the various components, and should not be
interpreted in a physical sense.
211
in equation (247). and an example of the graphical form of equation (247) is provided in Figure
69.
Q e .e u ( T c.d p ) ~ f i f e . d p ) (247)
As with the condensing unit, the compressor data can be used to convert Oe e „ into the
condensing capacity o f the evaporating unit ( Q c e u ).
0i . e,e i u ^
evaporating
unit
— ►
Te.dp
Figure 69: Graphical interpretation o f equation (247)
Comparison o f equation (246) and equation (247) implies that the condensing unit and the
evaporating unit can be matched to produce a system, whose equilibrium conditions are
determined when the expression in equation (248) is satisfied.
Q e .c u ^ c . d p ) Q e .e u i^ c .d p )
= 0 (248)
As with the conventional balancing technique, once Oe and Te dp are found, they can be used to
determine the corresponding Oc and Tc dp at equilibrium conditions through manipulation of
compressor performance data.
Table 27 lists the key stages for the conventional balancing technique, and the new ly developed
interim technique for ATG . Capacities for the respective stages are identified in terms o f the
dependent temperature variables, and the Table summarises the main differences between the two
techniques. Note that the 'and' term means that both variables are needed to define the capacity,
whereas the ‘or' term means that the variables are interchangeable so either can be used.
212
Table 27: Comparison o f capacity and temperature variables fo r balancing
For ATg = 0 K For ATG > 0 K

Component/system
Temp. Capacity Capacity
Temp, variables
variables variables variables
Compressor Te and Tc Q e.co m p ' Q c.com p T e.dP and Tc dp Q e .c o m p ' Q c.com p
Comp
Condenser Tc Q c .c o n d T c .dp Q c .c o n d
onent
Evaporator L Q e .e va p T e.dp and Tc dp Q e .e vo p
Condensing
L or Tc Q e .e u ■
>Q c .c /u T e .dP or Tc dp Q e .e i u • Q c .c /u
Sub unit
system Evaporating
- - T e.dP 0 f T c,dp Q e . e / u ' Q c .e iu
unit
System System L or Tc Q " Q c T e.dp ° r T c,dP Q e 'Q c
4.4 New balance technique for zeotrope with desuperheat and pressure loss
As discussed in Section 3.4.3, the heat exchanger pressure loss and compressor discharge
superheat/condenser desuperheat should be accounted for. It is important to first clarify how their
presence impacts on the logic of the balancing concept.
■AT,
lines of constant
TEl' 1n )
pressure
basic cycle (no

temp glide, no
press loss)
advanced cycle
(temp glide,
press loss and
e(aut)
disch. superheat)
Pei out)
Figure 70: Refrigeration cycle with pressure loss and superheat
213
Examination o f the effects o f Ap in heat exchangers showed that the corresponding change in
saturation pressure affects ATG . However, Ap across the condenser causes a difference between
inlet and outlet Tcdp. and consequently affects the interpretation o f Qe evap. To ensure
consistency. Tcdp determination of Oe es.ap must be linked to the correct reference state as used by
the condenser and compressor; this implies that the condenser Ap characteristic must become a
parameter in the balancing technique.
The second consideration is condenser A7 ^ , which was previously shown to be dictated by the
compressor performance. Observ ations from condenser simulations showed that A T ^ had a
significant effect on Oc cond (and A p ). It was shown that for a specific Tcdp(w,. the A
influences condenser Ap w ith the secondary effect o f reducing saturation temperature leading to
a change in ATa and Tc dp(out). In this respect the amount o f compressor A7 ^ must be
implicated in balancing when using a zeotrope.16
A refrigeration cycle has been plotted on a pressure-enthalpy chart in Figure 70, showing a basic
processes with evaporator and condenser Ap and compressor A . Ev aporator and condenser
Tjp that correspond to the inlet and outlet pressures are also shown. The pressure that the
refrigerant enters the condenser corresponds to a specific dew-point temperature ( Tc dp(in)). The
refrigerant cools, and reduces in pressure until it reaches a saturation point. During the
condensing process, the refrigerant temperature reduces as a result of the ATa , but also due to
further A p . At the end o f the condensing process the bubble-point is reached and (as subcooling
is neglected), this point is also Ttei.(w) . As discussed earlier TTEV(in) in the evaporator
characteristic w as substituted for the corresponding dew-point (equation 244); to ensure

consistency this reference point should ultimately be expressed in terms of Tc dpiin). To do this,
knowledge o f the condenser Ap is needed, which has to be identified from the condenser
characteristics. Since the evaporator rating is independent o f the condenser, the evaporator
capacity’ is initially7rated according to Tc dp(out). Following expansion, the refrigerant enters the
evaporator at Te(jn) , as determined by x e(in) and glide properties. However, this condition is o f no
16 It is noted that this parameter does not only affect systems with temperature glide, but also
those with pure refrigerants as the overall condenser capacity is changed regardless of the
existence o f temperature glide.
214
direct consequence to the balancing procedure, since it is already implicit within the evaporator
characteristic. Evaporation then follows where further Ap occurs, coupled with ATG and ATsh,
and the refrigerant eventually exits the evaporator at a superheated condition. The exit pressure
corresponds to a dew-point temperature which is designated Tedp(out).
1condenser
Figure 71: New condenser characteristic
In order to account for pressure losses and compressor discharge superheat in the condenser, the
definitions o f the condenser characteristic must first be modified. Since there are now two values
o f Tcdp corresponding to its inlet and outlet pressure, o f which the former was selected for the
reference condition. Thus taking equation (195) and ignoring ATx gives equation (249).
Q c .c o n d ~ f{^c.dp<w)'^Jsh ) (249)
The existence of the additional desuperheat variable, the condenser characteristic implies that the
condenser now exhibits several characteristic lines corresponding to the different values o f A7 ^
(see Figure 71).
T e,d p { o u t) 0 )
e,d p {o u t )
i compressor
e,dp(cnd)
T
x c,dp(in )
Figure 72: Discharge superheat characteristic
215
Condenser rating can only be based on fixed increments o f A a s the degree o f desuperheat
produced by a compressor at the rating stage is unknown. Therefore the compressor A m u s t
also be characterised, as previously given in equation (190). Since compressor Tc dp = Tcdp{in) of
the condenser and compressor Te dp = Te dp(oul) o f the evaporator, then equation (190) can be
rewritten as equation (250). This characteristic is shown graphically in Figure 72.
M d,h = f ( T , ,dp(out) »1 e.,dp(in) ) (250)
Similarly, the compressor capacity should also be defined with these new reference temperatures
so altering equations (188) and (189) gives equations (251) and (252) respectively.
Q e.co m p =f { r, ,dp{out) ’ r^ 'c,dp(in) ,) (251)
oc.cnmp = Jfir
\ te .d p (o u t) ’ c,dp{in) ) (252)
c.cond
( 1)
( 2) condenser
e, dp (out) (3 )J
Figure 73: M odified condenser characteristic
The next stage is to combine the condenser and compressor to form a condensing unit. However,
now that Oc cond is dependent on A T ^ this is not initially possible as seen with the dissimilar
variables when comparing equations (249) and (252). The solution to this problem is to utilise the
compressor A d a t a to convert condenser data rated for a specific A 7 ^ to capacities that
correspond instead to an equivalent evaporating temperature. In other words, for a fixed set o f
compressor operating temperatures, A c a n be determined, and the corresponding Qc cond
obtained for each o f the Tc dp(w) and Te dp[out) increments. From this, a new condenser
characteristic can be produced that is a direct function o f both the reference temperatures o f the
compressor. Thus equation (249) becomes equation (253) when Te dp(out) is substituted for A ;
216
the graphical interpretation of modified condenser characteristic is provided in Figure 73, and can
be compared with Figure 71.
Q c .c o n d f { T c.dp(in) ’ 1 '<?,dpiout) ) (253)
q A condensing unit
>(D
(2 ) compressor
condenser 1 (3)
Figure 74: Compressor and condenser balancing
Following this conversion, both the condenser characteristic (equation 253) and the compressor
condensing capacity characteristic (equation 252) use identical variables ( Qc cond, Tc dp(jn) and
e.dp(out) ). Now the modified condenser characteristic can be matched to the compressor, where
the balance points correspond to the same Te dp{out) o f each component (equation 254).
Q c .c u Q 'c .d p iin ) ) — f i ^ e . d p ( o u l ) ) (254)
The compressor and condenser characteristics and corresponding condensing unit characteristic
are shown in Figure 74. Equation (254) is an expression for this condensing unit characteristic
where the condensing temperature is implicit, and by default, A m u s t also be implicit.
' evaporator
(3)
Figure 75: Evaporator characteristic
217
Now with reference to the evaporator, by adopting equation (196) it is known that Oe ^,ap is a
function o f both 7,7E-r(l„) and Tedp(out), although as with the interim technique. TTEl,(in) becomes
Tcdp. However, because the condenser Ap is unknown, the existing evaporator characteristic is
not suitable because its current condensing reference temperature corresponds to the condenser
outlet pressure ( Tcdp(outl). For the time-being, the evaporator characteristic is redefined in
equation (255) and shown graphically in Figure 75.
,dp(out) 5 '^'c,dp(,out) ) (255)
Following the procedure for the interim technique, the next stage is to link the evaporator and
compressor to produce an evaporating unit. Comparison of equation (255) and the compressor
evaporating capacity characteristic (equation 251) reveals a disagreement with respect to the
reference condensing temperature. The compressor rating uses Tc dp{in), whereas the evaporator
rating assumes Tc dp(out). In order to account for this, two options are available: either the
evaporator or the condensing unit must be re-rated with respect to the condenser pressure drop.
These can be achieved by substituting the evaporator rating condition of Tcc,dp{out) with the
equivalent value o f Tc dp{w). or substituting the condensing unit rating condition Tc dpim) with the
equivalent value o f Tc dpiout), respectively for the condenser.
1 condenser
Figure 76: Condenser pressure loss characteristics
Ultimately, the favoured option is the former, since this will maintain consistency in the
component reference points, i.e., the dew-point corresponding to the pressure at the compressor
discharge. Initially, to determine the condenser pressure loss, data must be available. As
described in Section 5.3, Ap is implicit in the revised rating method and is therefore a function o f
218
Tc dp(m) and A . Accordingly, the equivalent Ap can be expressed in terms of the difference
between in inlet and outlet saturation temperatures, where it is influenced by A7 ^ (equation
256); a graphical representation is given in Figure 76.
c.dpôut) ~f { fie .d p ( in ) ■> ) (256)
Te,dp«m,)(2) condenser
Figure 77: Converted condenser pressure loss characteristics
In its present format this Ap data cannot be applied to the conversion of the evaporator data,
since the evaporator characteristic in independent of A . In equation (256), the absent variable
is Te dp(out), indicating that the condenser Ap characteristic needs to be converted from a function
o f A 7 ^ to Te dp(out). Following the earlier discussion on the conversion of condenser capacity,
the same logic can be applied here. It is known that for each set o f operating temperatures, the
compressor will produce a corresponding A . Therefore, the equivalent evaporating
temperature is determined from the compressor A7 ^ data at the specified Tc dp{in). Figure 77
illustrates the revised condenser pressure loss characteristic resulting from this conversion.
Tc,dp(m)0-) evaporator
T e.dp{out)
Figure 78: M odified evaporator characteristics
219
With this converted condenser Ap characteristic, the existing evaporator characteristic can be
modified so that Oe evap is expressed as a function of Tc dp{w) rather than Tc dpioul}. Equation
(255) becomes equation (257) and the graphical representation is as Figure 78.
Q e .e v a p ~f ^ c ,d p ( in ) ' ^ e,dp{out) ) (257)
This characteristic is now not only specific to the selected evaporator, but also specific to the
selected condenser and compressor. Now that Oe evap is a function of the same reference
temperatures as Oe comp (equation 251), it is simple to link the two to produce an evaporating unit
(equation 258).
Q e .e u ( T c ,d p U n )) = f e .dpio u t ) ) (258)
evaporating unit
0)
evaporator compressor
c,dp{in\ (3)
Figure 79: Balancing o f evaporator and compressor to form an evaporating unit
This balancing exercise is illustrated in Figure 79, where it is seen that the condensing
temperature becomes implicit in the evaporating unit characteristic, and indirectly, A .
evaporating unit
x balance point
condensing unit
e,dp(ont)
Figure 80: Evaporator balance point fo r condensing and evaporating unit
The final stage is to link the condensing unit and evaporating unit to obtain the overall system
balance points, and this can be carried out using condensing or evaporating capacity as a basis.
220
Using the former, Occ!u is a function o f Tc dp(w} (equation 254) can be converted to evaporating
capacity as a function o f Tedp{out) as explained previously for the conventional and interim
techniques, to give equation (259).
Q,C = ) (259)
Thus, the evaporating unit and condensing unit are balanced to provide the system equilibrium
evaporating temperatures and evaporating capacity (equation 260); a graphical interpretation o f
this is illustrated in Figure 80.
Qe.c u^c.dpiin)} Q'e.e ' u (^c.afp(i>i)) n /~%sr\\

— i—
~J jrVfe,dp(out) } —r
Jj iVf e.dp(out) )
Similarly, with these balance points known, the corresponding balance points for the condensing
temperature and condenser capacity can be determined (equation 261); this is also shown
graphically in Figure 81.
Qc.c uCê.dpiout)) Qc.e A Te.dp(out)) n

“7T
J VTC,dp(m) /1
--------71?—
J \ J c,dp{m) ) r (26I)
Subsequenth, values for Tc dpiout) and can be obtained.
condensing unit
balance point
Figure 81: Condenser balance point fo r condensing and evaporating unit
The key stages in the new balancing technique (revised for Ap and A ) are listed in Table 28,
with the corresponding temperature and capacity variables. The interim technique for ATG only
is also included in Table 28 for comparison. It is seen that the revised methodology is more
complex, since in addition to the evaporating unit, accounting for Ap and A d e m a n d s
221
additional stages for data conversions. Otherwise, the capacity and temperature variables are
essentially the same as for the interim technique.
4.5 Clarification of new procedures

A summary o f this new procedure with all relevant stages o f data manipulation is detailed in
Table 33 (Appendix 3) where the relevant equations are identified. The technique detailed here
presents a standardised procedure for balancing system components. It can be used for both pure
and zeotropic refrigerants. In addition, the final technique can be applied regardless o f whether
rating data is available for all o f these additional parameters since the relevant steps necessary to
account for a specific parameter can be neglected if the relevant data is unavailable.
Table 28: Comparison o f capacity and reference temperatures fo r balancing procedure
for ATg > 0 K. Ap = 0 Pa for ATg > 0 K, Ap > 0 Pa
Segment (interim technique) (full new technique)

Capacity Capacity
Temp, variables Temp, variables
variables variables
Q e .c o m p ' T e.dp{out) Q e.com p '

Compressor Te , d p ^ T cdp
Q e.com p T c .dp(in) and Q c.com p
Comp
Condenser T e.dp Q C T c .dP ■ AT^ Q c
onent
T c.dp(out)and
Evaporator Te.dp ™*Tc dp Qe Qe
Te.dp{out)
Condensing Te.dp ™Tcdp

Q e .d u ' Q e .d u *
T c.dpiiri) T g jjp ^ ^ t)
unit Q c .c - V Q c .d u
Sub
system
Evaporating Te.dp or Tc dp
Q e .e /u ’ Q e .e lu *
Tc,dp(iri) or T ! dp(out)
unit Q c .e /u Q c ,u
System System Te.dp or Tc dp Q e ’ Qc Tc,dp(in) or Tedp(OUt^ Q e ’ Qc
Finally, it is noted that the choice o f reference temperatures (i.e., Tcdp(m) and Te dp(out)) is
beneficial from a practical sense. When making field measurements, the easiest readings to take
tend to be suction and discharge pressures at the compressor. Neglecting Ap in suction and
discharge lines, these measurements enable dew-point temperatures to be estimated, thereby

providing a direct interpretation o f the real balance-points.
222
SECTION 5: ILLUSTRATIVE EXAMPLE OF NEW TECHNIQUES
5.1 Introduction
The intention o f this section is to work through the methodology by means o f a practical example.
There are three purposes for this example. The first is to assist the user in following the
technique; a practical example helps to convey its application to the user. Secondly, it is used to
validate the methodology; simulated component catalogue data was used, and the results of the
exercises are then checked against component simulation data for the conditions at the respective
balance points. Capacities from the balancing exercise should equal those obtained from the
direct simulation at balance temperatures if the proposed technique is valid. Lastly, the accuracy
o f the new method can be compared against existing techniques to establish whether the new'
technique provides a benefit over the existing techniques; a comparison can be made against the
results obtained using the conventional approach where references temperatures are interpreted as
suggested in the literature. For example, Hundy (2000) and Murphy et al (1998) describe using
both dew-point conditions and mid-point conditions for compressor rating, and Atkins (1999)
uses mid-point for evaporator and condenser rating. The definition of mid-point temperature for
the condenser ( Tc mp) and evaporator ( T ) are according to equations (262) and (263)
respectively.
(262)
(263)
By substituting T+ or Tmp for pure refrigerant evaporating and condensing temperatures in the
component performance data, the exercise can be worked through using the conventional
balancing technique. Comparison o f the results against those from the new technique can be used
to judge the improvement in accuracy.
5.2 Generation of component data

Eight examples were used to cover a range of system characteristics that the new balancing
technique was intended to account for; specifically small and large ATG, heat exchanger A p , and
a range of AT(in). Table 29 shows two different heat exchanger combinations were used with
small and large ATG (systems 1 and 2, and 3 and 4), each with (a) small size and (b) large size
compressor. Condenser and evaporator Ap w ere varied by tube length and number of refrigerant
223
circuits. For each component design, corresponding models were used to generate fictitious
performance data as for a catalogue.
Table 29: Specific input data for calculation o f fictitious rating data
System 1 System 2 System 3 System 4

Component Variable (high ATg . (high ATg . (low ATg , (low ATg ,
high Ap) lowr A p) high Ap) low Ap)
Temp, glide at
Refrigerant 10 K 10 K 5K 5K
0°C
Comprcsso (a) Swept volume 15 m3h_1 15 n ^ h '1 15 m3 h'1 15 m3h']

r (b) Swept volume 35 m3 h'1 35 m3 h'1 35 m3 h'1 35 m3 h 1
Desuperheat 30 K - 90 K 30 K - 90 K 30 K - 90 K 30 K - 90 K
Condenser No. circuits 3 4 3 4
Circuit length 8.0 m 3.5 m 8.0 m 3.5 m
No. circuits 6 12 6 12
Evaporator
Circuit length 3.0 m 2.0 m 3.0 m 2.0 m
The refrigerant thermodynamic and transport properties were calculated for a mixture o f 50%
R290 and 50% R600a by mass, although an artificially imposed glide was used as explained in
Section 3. For the two refrigerant characteristics, the glide at Tip = 0°C was 5 K, and 10 K and
both had a glide shape with index nG = 3.
Compressor rating data was based on 5 K evaporator superheat and no subcooling. Both
compressors had a clearance ratio o f 0.15. Similarly, the condenser rating data was also based on
0 K subcooling, and HTF inlet temperature was +20°C, for a water flow7rate o f 0.4 kg s'1. The
inside tube, containing refrigerant had a diameter o f 0.010 m, and the surrounding tube
containing HTF w as 0.025 m diameter and total tube length and the number o f circuits differed.
The evaporator used a similar construction, with an inside refrigerant tube diameter of 0.015 m,
and the HTF again passed through a 0.025 m tube. The HTF mass fknv rate was 0.35 kg s'1, and
entered at +15°C, and rating data was again based on 5 K superheat, consistent with the
compressor. Neither condenser nor evaporator used extended heat transfer surfaces. Lastly, it
was assumed that all components were directly connected, to negate the possibility o f temperature
and pressure changes within interconnecting piping and ancillary components.
224
5.3 Illustration of worked example
The worked example was carried out using the data for System 1(a), although the results for all
eight systems are provided.
Stage 1: Compressor characteristic data

Compressor condensing capacities, and discharge superheat catalogue data was obtained (Figure
82 and Figure 83). It is noted that the reference temperatures are not designated for {in) or {out)
conditions, which is because they do not necessarily refer to the heat exchangers. Logically
though, since the compressor suction is connected to the evaporator outlet, its Te dp = Te djKout) of
the evaporator, and similarly because the compressor discharge in connected to the condenser
inlet its Tcdp = Tc dp(m) of the condenser.
10
Te dp - -40 C
— 7 U = -3 < r c
8
T,dp = *20^0
—
6 ---------- T'.dP = 0C
— ----------------- —
4
- - - -r- ■ - ------- --------------------------------------- r
0 30 40 50 60
-40 -30 -20 -10 0 10 Tc.ap (°C)
Te,* (°C)
Figure 82: Compressor capacity, against Figure 83: Compressor discharge superheat,
evaporating dew-point temperature fo r a against condensing dew-point fo r a series o f
series o f condensing dew-point temperatures evaporating dew-point temperatures
Stage 2: Condenser characteristic data

For the selected condenser, Qc capacity and equivalent Ap catalogue data is obtained (Figure 84
and Figure 85). The Ap data is presented in terms of inlet and outlet dew-point temperatures
( Tc.dpirn) *tnd T c dpôuf) ).
225
Stage 3: Evaporator data
For the selected evaporator, Oe evap catalogue data is obtained as a function of Te dp<out) for a
series of Tc dpioul) (Figure 89).
12
A7d* =10 K AT^, = 10K
A =50K AT^. = 5 0 K
A 7^=90K A 7^= 9 0 K
v AT^j. = 130K
A 7^=130K
30 40 50 60
T c.dptmi (°C ) Tc.dp(m) (°C )
Figure 84: Condenser capacity. against Figure 85: Condenser outlet dew-point
condensing dew-point temperature fo r a series temperature, against condenser inlet dew-
o f discharge superheats point temperature fo r a series o f desuperheats
Stage 4: Condenser characteristic data matched to compressor superheat

The condenser capacity is to be matched to the compressor, but reference to Oc comp (Figure 82)
and Oc cond (Figure 84) are a function o f different variables. Therefore, the condenser
characteristic lines in must be modified so that the capacity is also given as a function of
Te.dp(out) (rather than A7 ^ ) . This is achieved through use o f the compressor A d a t a (Figure
83), so for a given condenser data-point in Figure 84, the corresponding Te dp{outj is taken from
Figure 83.
Once all data-points have been converted, a new condenser characteristic is produced where
Q c ,c o n d is presented against Tc dp{in) for a series o f Te dp(out) (Figure 86), which requires
interpolation. It is noted that A i s implicit in the new condenser characteristic. This same
conversion procedure is also carried out for the condenser pressure drop characteristic, since the
data is needed for a series of Tedp(out) instead o f A ( F i g u r e 87).
226
ip (out) 40°C T e ^ io u t) = - 40^
‘ T e ,d p (o u t ) = -30°C Teôut) =-30°C
•Te^ {out)=-20°C ?e,dp(pul) ~2(j C
1 50
- T e, M o u t ) = - \ 0 ° C o e,cf?(out)
T-e,dp(out) = 0 ° C
‘ 'Te, dp(out)
30 4 0 50 60
T c.dpi m) ( ° C ) c ,d p \ in )
Figure 86: Condenser capacity against Figure 87: Condenser outlet dew-point
condensing dew-point temperature, fo r a temperature, against condensing inlet dew-
series o f evaporating temperatures point temperature fo r a series o f evaporating
temperatures
Stage 5: Condensing unit characteristic

Now that both Oc cW and Oc comp data are a function of the same variables ( Tedp(out) and
Tc,dp{in) X they can be matched to provide a condensing unit characteristic. The condenser
characteristic is superimposed on the compressor data, although it is converted to be a function o f

Te,dpiout) a series o f Tc dp(in). Where the component characteristics of the same Tc dp(w)
intersects, indicates the condensing capacity o f the condensing unit (Figure 88).
12
0 -I 1---------------- 1---------------- 1---------------- 1---------------- i
-40 -30 -20 -10 0 10

T e .d p ( o u t ) (°C)
Figure 88: Compressor characteristic superimposed on new condenser characteristic to obtain

condensing unit characteristic
227
Stage 6: Evaporator characteristic matched to condenser inlet
The next two components to be balanced are the evaporator and compressor. However, before
this can be achieved, the evaporator inlet condition has to be determined. Given that Qeevap is
presently based on Tc dp(out) (i.e., into the TEV, ~ , Figure 89), whilst Oecomp is based on
Tcdp = Tc dp{in) (Figure 82), the evaporator characteristic needs to be adjusted so that reference
states are consistent.
The data within the modified condenser pressure drop characteristic (Figure 87) is therefore used
to convert the evaporator characteristic from Tc dp(oul) to Tc dp(w]. This is achieved as follows.
From the evaporator data (Figure 89), a Tc dp(out) and Te dp{out) coordinate is selected and applied
to the converted condenser pressure drop data (Figure 87) for which the corresponding Tc dp(w) is
determined. For the chosen Tc dp(out) and Te dp(out), the Qe e%,ap value from Figure 89 is plotted
against the original Te dp(out) and the identified Tc dp{m). This procedure is followed for the entire
evaporator data set, and Figure 90 shows the resulting modified evaporator characteristic.
c,dp(out) -3 0 ° C T c. m n ) - 3 0 ° G
Tcdpâut) = 4 0 °C I'c.dpUn) = 4 0 ° C :
T c, M oui) = 50 °C
T c,d p (p u t) ~ 60°C
Oi
-40 -30 -20 -10 10 -40 -30 -20 -10 0 10

Te,dp(out) ( ° C ) Te.dp(out) ( ° C )
Figure 89: Evaporator capacity against Figure 90: Evaporator characteristic

evaporating dew-point temperature fo r a corrected fo r condenser inlet
series o f condensing outlet dew-point
temperatures
228
Stage 7: Evaporating unit characteristic
As with the condenser, the evaporator is matched with the compressor to form the evaporating
unit. Both compressor evaporating capacity characteristic (Figure 82) and the modified
evaporator characteristic (Figure 90) are now a function o f the same reference states ( Tc dp(jn) and
Te dp(oun)- and therefore they can be linked. Figure 91 shows Qecomp superimposed on Qe evap and
the evaporating unit characteristic corresponds to the intersection of their respective Tc dp(m).
t c.M u> = 30°C
Tcdp{,.> =40°C
= 50°C / /
= 6 0 °C
-O -evapu n i
e.evap
-4 0 -3 0 -20 -10 10
T e,dp(out) ( ° C )
Figure 91: Compressor characteristic superimposed on new evaporator characteristic to obtain

evaporating unit characteristic
Stage 8: Convert condensing and evaporating unit capacity characteristic

Once the characteristic for both condensing and evaporating units are obtained, the balance points
for the total system can be determined. This is achieved by identifying the intercept o f the units'
capacities. However, it is noted that the condensing unit characteristic is currently in the form of
Qc c/u over a range of Te dp(out) (with implicit Tc dp(w)) (Figure 88), whilst evaporating unit
characteristic is Qe e/u a function o f Te dp(out) (Figure 91). To balance the system, the capacity of
one unit must be converted to the capacity of the other. Using the compressor data, the
condensing unit characteristic is converted to the equivalent evaporating capacity at the
corresponding Te dp{out) and Tc dp{m). Similarly, the opposite conversion can also be made, where
Qe e/u is converted to Qc e u, using the compressor characteristic data. It is noted that this is the
same process is required for the conventional balancing process.
229
Stage 9: Balance condensing unit with evaporating unit
The final stage is relatively straightforward. The evaporating unit and condensing unit are
matched twice: once for Oc as a function o f Tcdp{in), and for Oe as a function of Tedp{out}. Thus,
the intercepts o f these provide the system equilibrium Oc and Oe and corresponding Tc dp{in) and
Tedp(out) (Figure 92 and Figure 93). The numerical values of these balance points are provided in
Appendix 4, Table 34.
10 1 conduril, QCClU cond ink, (?e.c «

8 - evapun<Qcelu evap unit, Q e,e /u
6 -
O balance point O balance point
6 -
£ ■ o
4 - &
2 -
2 -
0 -
3 0 4 0 50 60 •1 5 -10
Tc,dp{w\) ( ° Q Te,dp(out) (°C)
Figure 92: System balance condensing dew- Figure 93: System balance evaporating dew-
point temperature point temperature
5.4 Comparison of results

The accuracy o f the new technique was assessed by comparing the results o f the eight system
balance exercises, against the output o f the component models based on Tc dp{in) and Te dp(out) . In
addition, the conventional procedure for system balancing was also used to find equilibrium
points in the normal way, where the dew-point and mid-point temperatures were used as reference
states (Section 5.1). The results are listed in Appendix 4 (Table 34). An overview o f the “error”
associated with each approach is presented in Figure 94, for the systems using (a) small size
compressor and (b) large size compressor. The error is calculated as the ratio of capacity
returned by the respective technique to the modelled capacity (which is taken as the correct
value).
In Figure 94, it can be seen that the new7balancing technique generally concurs with the modelled
capacities to within ±1% for the condenser, and within ±2% for the evaporator. Whilst the new
230
technique is deemed to be conceptually correct, minor disagreement with the modelled values is
expected, due to successive linear interpolation between performance data-points. With reference
to Figure 88, Figure 91, Figure 92, and Figure 93, it is observed that the gradient of the
evaporating capacity characteristic has more curvature than the condensing characteristic, which
is consistent with the greater error associated with the evaporator capacity due to linear
interpolation being less accurate. Nevertheless, given that new technique produces capacities at
system equilibrium conditions consistent with those of the model over the broad range of system
configurations suggests that it achieves its purpose.
20 20
evaporator e\aporator
ii condenser K condenser
15 15
5 5
0 0
Svstem3
Figure 94: Error in Oe and Oc at balance points with (a) small size compressor, and (b) large
size compressor fo r conventional and new techniques
Figure 94 show s that unlike results from the new technique, capacities estimated with the
conventional method using or Tmp are always over-predicted compared to the models. Also,
the evaporating temperature is always higher than that predicted by the new technique (see Table
34. Appendix 4). A higher evaporating temperature means a smaller ATe(in), and so lower Oe
would normally be expected according to the evaporator characteristic: this is contrary to results
of the balancing exercise. The fact that the conventional method overestimates Oe for a smaller
ATe{in) demonstrates that the error is actually more severe than implied by Figure 94.
Furthermore, the overestimation of capacity is worse w here mid-point temperature is used

considering that ATe(/w) based on Te mp implies a larger Oe than when calculated with Te dp.
Ultimately, use of the conventional balancing technique using either r * or Tmp is expected to
dramatically underestimate the capability of the system to meet the heat load.
231
With respect to the condenser, over-prediction o f Oc is proportionally less than w ith Oe ,
although the difference between equilibrium condensing temperatures are negligible across all
approaches. Since the errors associated with the condenser are smaller, this suggests that most
focus should be directed towards accurate evaporator characterisation. Finally, it was noted that
under no circumstances does the use o f the conventional balancing technique, using T* or Tmp
provide better accuracy than the new technique. Neither can a preference be made in favour of
adoption o f mid-point or dew-point when used in the conventional balancing technique.
As mentioned earlier, the example systems (Table 29) represent varying sizes o f ATG, A p ,
ATe(//j) and A T . It is useful to review the results in respect o f these differences, to identify
any aspects demanding further attention. Observation of errors from the new technique does not
show any distinct trends across any o f the range o f variables. For the conventional technique,
comparison o f the results for high and low’ ATG do not reveal any consistent differences either.
Generally, the greater errors are for Oe when ATG is large, depending upon whether T+ or T
is employed, although low ATG tends to result in a greater error for Oc . The systems with high
Ap consistently (except one) produce significantly greater error than those with low A p , since a
larger Ap causes a wider shift in saturation temperatures and therefore over-prediction o f
capacities; consistently, errors were greater with the large size compressor, which implicitly
causes higher A p . A higher refrigerant mass flow equates to an increase in capacity, thereby
producing larger Ap and simultaneously depressing Te dp and elevating Tc dp. Quantitatively, the
errors in capacity generated in these examples range from about ±1% to +15% for Oc and ±20%
for Oe , with the largest errors being associated with the conventional technique using or Tmp
for a system with large size compressor. The over-prediction of capacity observed in both
approaches o f dew- and mid-point using the conventional technique is consistent with the fact that
it does not account for these parameters ( ATG, Ap and A ) w hen operating under non
standard conditions (e.g., reduction in capacity with Ap and variable A ). With this it is
evident that an improvement in accuracy o f prediction of equilibrium conditions will always be

achieved when using the new technique.
232
SECTION 6: FINAL SUMMARY AND CONCLUSION
6.1 Summary of work

The objective of this study was primarily to develop a means o f integrating the characteristics of
zeotropic refrigerants (which exhibit a temperature glide) and associated effects o f pressure drop
and sensible heat transfer in heat exchangers, into design activities within the component-system
chain. Several stages within this chain necessitate the generation, manipulation and interpretation
o f individual component performance data to enable correct selection and design o f the system.
O f particular relevance to this are the performance testing/rating, catalogue data, selection and
system balancing stages.
Particular attention w as paid to the behaviour o f condenser and evaporator performance when
using a zeotrope. Results from a detailed performance simulation were used to clarify the
implications o f zeotropes. Accordingly, new rating methods and selection o f suitable reference
state temperatures were proposed. In terms of system balancing, the conventional technique was
adapted to produce a new technique. Finally, use o f the new technique was illustrated by means
o f a w orked example.
6.2 Key findings and originality

The literature review revealed there were virtually no studies relating to component
characterisation and system balancing when using zeotropes. The review' o f component rating
standards and catalogues showed no consistency in the choice o f suitable reference states. An
assessment o f current methods o f performance rating and catalogue data indicated that there w ere
inconsistencies between these methods and the parameters necessary to handle zeotropic
refrigerants correctly. In particular, it was established that the condenser and evaporator were
affected in terms o f choice o f rating reference states, as w as the general logic for the basis of
rating and balancing.
It was found that the conventional definition o f approach temperature difference was not valid
since the existence o f the glide prevents the saturated refrigerant inlet temperature from reaching
HTF inlet temperature, which is contrary to the premise for standardised methods, thus the
necessity for a revised characterisation method. Analysis of evaporator performance also
identified the dependency o f its capacity' on the inlet condition of the refrigerant due to the
intersection o f the temperature glide.
233
Observation o f simulation results confirmed that as reffigerant-HTF inlet temperature difference
o f a condenser or evaporator approached that o f the temperature glide, exchanger capacity would
theoretically converge on zero (despite the existence of a temperature difference). It was also
found that capacity was also largely affected by the shape (up to ±200%) and size of the glide (up
to ±100%) especially at small approach temperature differences. Simulations also showed
condenser capacities to have a strong dependency on desuperheat (30% at 30 K) and this
identified the need to include it within the condenser rating method. Pressure loss through
exchangers was also found to affect component rating due to the resulting effect on reference
state temperature (±20% at high approach temperature differences), and it was concluded that it
should be the condensing dew -point temperature corresponding to the condenser inlet pressure,
and the evaporating dew-point temperature corresponding to the evaporator outlet pressure.
The revised balancing technique was adopted to account for the various parameters associated
with a zeotrope identified to affect component performance. The new technique was compared
with the existing approaches (using mid-point and dew-point) where up to 20% over-prediction in
component capacity at the balance conditions was observed. These findings demonstrate the
improved accuracy associated with the revised techniques.
Various alterations to the production, use, and manipulation o f performance data within the
component-system have been proposed, and are detailed below7, and summarised in Table 30.
- Considerations fo r component performance testing bating. At the outset, the performance

testing o f the component should ensure that it accounts for all the parameters relevant to
latter stages o f the component-system chain. These parameters are detailed in Table 30
(Items 1, 2 and 3).
- Considerations fo r component catalogue data. The catalogue data should provide capacity7
data according to Table 30 (Item 1). Whether a range o f data is provided, or just at rating
conditions with correction factors, it must be based on the correct reference temperature as
detailed in Table 30 (Item 2). Options exist for the characterisation, or extension o f standard
rating data: it can be presented for an entire range o f conditions, or generated from a
characterisation method such as those proposed earlier. These options are detailed for the
respective components in Table 30 (Item 4).
- Considerations fo r component selection. Guidelines for correct component selection should

be provided in catalogue information. Components should be selected according to the correct
reference temperatures and all corresponding performance parameters in Table 30 (Item 3).
234
- Considerations fo r system balancing. To ensure accurate balancing, the new technique
should be employed. Component data according to Table 30 (Item 5) should be available to
complete the full system balance.
It is noted that the new methodologies are more complex than those found in the literature, and
therefore working through the relevant procedures would be a more time consuming.
Table 30: Summary o f usable data
Item Compressor TEV Condenser Evaporator
1. Capacity C?e.comp ' Q c.co m p Q e ,T E l' CPc.cond Q e,evap
Te,dp(out) ?
2. Reference Te ,d p (o u t) ' 1Tc.dpiih) T*e,dp(out)
temperature ■ T c ,dp(iri)
•T*c,dp(out)
3. Other data
A T d s* A ^ d s h - T c dp{out)
required
4. Character Double-quadratic Full data, rating Full data, rating
Equation (81) and
isation equation, formula, correction formula, correction
correction factors
method correction factors factors factors
Q e.co m p VS.
Te ,d p (o u t) ' Tc.dp(tn)
Q c .c o n d ' S
Q e .e va p VS
5. Data O c,comp vs T c ,d p {in )i ^ 7 'dsh •>
required for None Te ,d p (o u t)'
balancing Te.dp(out) ? -T*c.dp(irt) T c,dp(in) VS
Tc,dp(out)
Tc. dp( out)
AT*h vs-
Te ,d p (o u t)' Tc.dp(in)
The importance o f the present study can be considered with respect to the fact that the majority of
new refrigerants are zeotropic mixtures, which are gradually replacing the existing pure and
azeotropic mixtures. It was shown in Part 1 that close matching o f a systems cooling capacity
with the application heat load improves efficiency, which is becoming an increasingly common
prerequisite for new equipment due to the cost of energy, and environmental legislation. Given
that the new methodologies assist with better determination o f system capacity, a closer match
with the application can be made, thereby promoting better efficiency.

Several aspects were neglected in the development of the new methodologies, which could be
incorporated to improve applicability. These are explained below.
235
- Expansion devices. Other expansion devices, such as capillary tube or short tube orifices
may be non-adiabatic and, and its capacity does not automatically adjust according to the
demand. To include this type of expansion device within the balancing technique requires
additional stages and manipulation o f data, as well as a suitable rating method.
- Ancillary components. The new balancing technique assumes that compressor, TEV,
condenser and evaporator are directly connected. However, since most systems employ
interconnecting piping, a variety o f valves, liquid receivers and other components, their
influence on reference state temperatures (due to pressure losses) and rated capacities o f the
main components (due to heat transfer) should be accounted for.
- Composition shift. Any zeotropic refrigerant will be subject to a variation in circulating
composition, and this affects refrigerant properties and consequently, capacities, glide
characteristics and definitions o f reference state. In addition to the characteristics of the
mixture, the degree o f composition shift is largely dependent upon system configuration,
which makes it difficult to account for at the component rating and balancing stages.
Nevertheless, integration of the influence of composition shift into both component
characterisation and system balancing would provide additional benefits.
- Influence o f oil. When the oil and refrigerant are mutually soluble, there can be an impact on
refrigerant behaviour and properties. A combination o f a zeotropic refrigerant and soluble oil
complicate the issues, due to preferential solubility o f the different refrigerant components
within the oil. Evidently, to account for oil effects within the rating and balancing techniques
would provide a more comprehensive approach.
It would be advantageous to produce a computational code to automate the manipulation o f rating

data and ultimately determine system equilibrium conditions. As such, inclusion of the influence
o f ancillary components w ould be simplified (for the user). More importantly, the code could be
extended to “re-rate” various components that are currently presented in catalogues with
insufficient detail.
236
REFEREN CES
ASHRAE 17-1986 (RA 1990). Method o f Testing for Capacity Rating of Thermostatic
Refrigerant Expansion Valves. American Society o f heating, Refrigeration and Air Conditioning
Engineers, Atlanta, USA, 1986.
ASHRAE 23-1993. Methods o f testing for rating positive displacement refrigerant compressors
and condensing units. American Society o f heating. Refrigeration and Air Conditioning
Engineers, Atlanta, USA, 1986.
ASHRAE 25-1990. Methods o f testing forced convection and natural convection air coolers for
refrigeration. American Society- o f heating, Refrigeration and Air Conditioning Engineers,
Atlanta, USA, 1990.
ASHRAE. Fundamentals Handbook (SI). American Soc. Heating, Refrigeration and Air-
conditioning Engineers, Atlanta, USA, 2001.
ASHRAE. Refrigeration Handbook (SI), Chapter 44: Component balancing in refrigeration

systems. American Soc. Heating, Refrig. Air-Condition. Eng., Atlanta, USA, 1998.
Atkins. N. Personal Communication. Searle Manufacturing. June, 1999.
Bitzer. Catalogue data for .2-Generation reciprocating compressors, K P-100-3. Undated.
Bivens, D. B., Yokozeki, A. Heat transfer o f zeotropic refrigerant mixtures. Proc. IEA Heat
Pump Conf. "Heat Pumps for energy Efficiency and Environmental progress", Elsevier Science,
Netherlands, 1993.
Bobbo, S., Camporese, R., Cortella, G., Fomasieri, E. Theoretical evaluation of the performance
o f zeotropic mixtures in refrigerating cycles. Proc. 19th Int. Cong. Refrig., HR, The Hague,
Netherlands, 1995.
Castle, T. P., Richardson, R. N., Ritter, T. J. Thermostatic valve control using a non-azeotropic
refrigerant isobutane/propane mixture. Int. J. Refrig., Vol. 22, No. 4, pp. 313-318, 1999
Chen, J., Kruse, H. Calculating circulation concentration of zeotropic refrigerant mixtures. Int. J.
heating, Ventilating, Air-Conditioning and Refrigerating Research, Vol. 1, No. 3, July 1995,
ASHRAE, Atlanta, USA.
Colboume, D., Suen, K. O. A new rating method for heat exchangers using refrigerant with
temperature glide. Proc. Int. Cong. Refrig., HR, Sydney, 1999.
Colboume, D., Rajapaksha, L., Suen, K. O. System balancing for zeotropic refrigerants. Proc.
Int. Refrig. Conf at Purdue, USA. 2002.
237
Colboume, D., Suen, K. O. Advanced system balancing for pure and mixture refrigerants. Proc.
Inst. Refrig., 3-1, London. 2001.
Conklin, J. C., Vineyard E. A. Heat exchanger thermal performance for two non-azeotropic
refrigerant mixtures. Proc. 18th HR Int. Congr. Refrig., Montr., Vol. 3, pp. 1269-1274. 1991.
Danfoss. TN and TEN-type expansion valve data. Danfoss UK Ltd, London, Undated.
Dossatt, R. J. Principles o f refrigeration, Chapter 13: System equilibrium and cycling controls.
Prentice-Hall Inc., 4th Ed, pp 258-261, USA, 1991.
EN 328: 1993. Heat exchangers - test procedures for establishing performance o f unit air coolers
for refrigeration. British Standards Institution, London, 1993.
EN 1117: 1994. Heat exchangers - Liquid cooled refrigerant condensers - test procedures for
establishing the performance. British Standards Institution, London, 1994.
EN 1118: 1994. Heat exchangers - Refrigerant cooled liquid coolers - test procedures for
establishing the performance. British Standards Institution, London, 1994.
Gaibel, J. A., Chato, J. C., Dobson, M. K., Ponchner, M., Kenny, P. J., Shimon, R. L.,
Villaneuva, T. C., Rhines, N. L., Sweeney, K. A., Allen, D. G., Hershberger, T. T. Condensation
o f a 50/50 blend of R32/R125 in horizontal tube with and without oil. Air Conditioning and
Refrigeration Centre report ACRC TR-56, University o f Illinois at Urbana Champaign, May
1994.
Gosney, W. B. Principles o f refrigeration. Cambridge University Press, UK. 1982.
Gosney, W. B., Suen, K. O. Notes on component matching and system integration. Unpublished
manuscript, 2001.
Granyrd, E., Conklin, J. C. Thermal performance analysis for heat exchangers using non-
azeotropic refrigerant mixtures. Heat Transfer in Advanced Energy systems, HTD-Vol. 151,
AES-Vol. 18, pp. 25-32, American Soc. Mech. Eng., New York, USA, 1990.
Haselden, G. G., Klimek, L. An experimental study o f the use of mixed refrigerants for non-
isothermal refrigeration. Proc. Inst. Refrig., Vol. 54, pp 129-148, London, 1957.
Hewitt, G. F. Heat transfer including evaporation, condensing and mixed refrigerants. Proc. Conf
Inst. Refrig., London, November 1999.
Holladay, W. L. A rating method for the auxiliary desuperheater-condenser. Proc. ASHRAE

Symposium, paper no. 2845, Vol. 90, part 2A, pp. 302-313, American Soc. Heating, Refrig. Air-
Condition. Eng., Atlanta, USA, 1984.
Hundy, G., Vittal, R. Compressor performance definition for refrigerants with glide. Proc. Int.
Refrig. Conf at Purdue, pp. 415-424, USA, 2000.
Hundy, G. Wise-up to the impact of glide on performance. Refrig, and Air Condition., EMAP,
London. January 2000.
Incroprea, F., De Witt, D. Fundamentals of Heat and Mass Transfer. 3rd Ed., Wiley, Singapore,
1990.
ISO 817: 2005. Refrigerants — Designation and Safely Classification. British Standards
Institution, London. 2005.
ISO 917: 1989. Refrigerant compressors - measurement o f performance data. British Standards
Institution. London, 1989.
ISO 9309: 1996. Refrigerant compressors - Presentation o f performance data. British Standards
Institution, London, 1996.
Janssen, M.. Beks, P., Hermetic compressor performance evaluation applying a zeotropic
hydrocarbon blend instead of CFC-12. Proc. HR Conf. Natural Working Fluids, pp. 217-224,
Purdue University, USA, 2000.
Jung, D., Radermacher, R. Prediction of evaporation heat transfer coefficient and pressure drop
o f refrigerant mixtures. Proc. Int. J. Refrig., Vol. 16, No. 5, pp 330-338, 1993.
Jung, D ., Radermacher, R. Prediction of evaporation heat transfer coefficient and pressure drop
o f refrigerant mixtures in horizontal tubes. Proc. Int. J. Refrig., Vol. 16, No. 3, pp 201-208,
1993.
Jung, D., Radermacher, R. Prediction of pressure drop during horizontal annular flow boiling of
pure and mixed refrigerants. Int. J. Heat and Mass Transfer, Vol. 32, No. 12, pp2435-2446,
1989.
Kayansavan, N. Heat transfer characterisation o f plate fin-tube heat exchangers. Int. J. Refrig.,
Vol. 17, No. 1, pp. 49-57, 1994.
Kenny, P. J., Chato, J. C., Dobson, M. K., Wattelet, J. P., Gaibel, J. A., Ponchner, M., Shimon,
R. L., Villaneuva, T. C., Rhines, N. L., Sweeney, K. A., Allen, D. G., Hershberger, T. T.
Condensation o f a zeotropic refrigerant R32/R125/R134a (23%/25%/52%) in horizontal tube.
Air Conditioning and Refrigeration Centre report ACRC TR-62, University o f Illinois at Urbana
Champaign, July 1994.
239
Lemmon, E. W., McLinden, M. O., Huber, M. L. NIST Reference Fluid Thermodynamic and
Transport Properties - REFPROP. Version 7.0. National Institute of Standards and Technology,
Boulder, USA. 2002.
Liu, X. LMTD applications in two-phase heat transfer and two-phase heat exchangers. Proc. Int.
Refrig. Conf. at Purdue, pp. 145-150, USA, 1996.
Mulroy, W. J., Domanski, P. A., Didion, D. A. Glide matching with binary and ternary' zeotropic
refrigerant mixtures. Part 1: an experimental study. Int. J. Refrig., Vol. 17, No. 4, pp. 220-225,
1994.
Murphy, F. T., Corr, S., Morrison, J. D. Influence of refrigerant properties on standard

compressor rating figures with particular reference to standard superheat conditions and
refrigerant temperature glide. Proc. Inst. Mech. Eng. Seminar: "Design selection and operation of
refrigerator and heat pump compressors." London, November 1998.
Ntalles, N. A Study o f temperature glide effect of the non-azeotropic refrigerants. MSc.

Dissertation, University College London, 1997.
Page. A. O. The use of personal computers in refrigeration system design and analysis. Proc.
Inst. Refrig., pp. 34-42, London, 1988-1989.
Popovic, P.. Shapiro, H. N. A semi-empirical method for modelling a reciprocating compressor in

refrigeration systems. Proc. ASHRAE Transactions, paper no. 3912, pp. 367-382, American
Soc. Heating, Refrig. Air-Condition. Eng., Atlanta, USA, 1995.
Prestcold. PL, K, L, LG and R-type compressor data. Copeland Corporation Ltd, Reading,
Undated.
Rabehl, R. J., Mitchell, J. W., Beckman, W. A. Parameter estimation and the use o f catalogue
data in modelling heat exchanger coils. Int. J. Heating, Ventilating, Air-Conditioning and
Refrigerating Research, Vol. 5, No. 1, ASHRAE, Atlanta, USA, January' 1999.
Searle. M DR-t\pe air-cooled condenser data. KS and KM-tvpe unit coolers. Searle
Manufacturing Ltd, Fareham, Undated.
Shao, D. W., Granryd, E. G. Flow’ pattern, heat transfer and pressure drop in flow' condensation.
Part II: Zeotropic Refrigerant Mixtures (NARMs). Int. J. Heating, Ventilating, Air-Conditioning
and Refrigerating Research, Vol. 6, No. 2, April 2000, ASHRAE, Atlanta, USA.
Shao, D. W., Gramy d, E. Experimental and theoretical study on flow condensation with non-
azeotropic refrigerant mixtures o f R32/R134a. Int. J. Refrig., Vol. 21, No. 3, 1998.
240
Shin. J. Y., Kim, M. S., Ro, S. T. Experimental study on forced convective boiling heat transfer
o f pure refrigerants and refrigerant mixtures in a horizontal tube. Int. J. Refrig., Vol. 20, No. 4,
pp. 267-275, 1997.
Stephan, K. Two-phase heat exchanger for new refrigerants and their mixtures. Int. J. Refrig.,
Vol. 18, No. 3, pp. 198-209, 1995.
Stoecker, W. F., Jones, J. W. Refrigeration and air conditioning. Chapter 14: Vapour-
compression-system analysis. 2nd ed., McGraw-Hill Book Co., pp 281-295, Singapore, 1982.
Stoneham, H. G., Saluja, S. N., Dunn, A. Heat transfer and pressure drop characteristics of
refrigerants evaporating inside horizontal tubes. Proc. Inst. Refrig., pp. 52-60, London, 1979-
1980.
Suen, K. O. Personal communication. London. 2000.
Sumida, Y., Tanaka, N., Okazaki, T. Prediction o f the circulating composition of a zeotropic
mixture in a refrigerant cycle. Proc. 19th Int. Cong. Refrig., HR, The Hague, Netherlands, 1995.
Trane Air conditioning Company. Trane Air Conditioning Manual, Chapter VI: Refrigeration
theory, reciprocating compressors and cycle components. McGill/Jensen, pp 155-159, USA,
1988.
Trott, A. R. The compilation and interpretation o f catalogue data with simple mathematical
models. Proc. Inst. Refrig., pp. 58-68, London, 1981.
UNEP. 2002 Report of the Technical Options Committee on Refrigeration, Air Conditioning and
Heat Pumps. United Nations Environment Programme, Nairobi, 2002.
Venkatarathnam, G., Mokashi, G., Murthv, S. S. Occurrence o f pinch-points in condensers and

evaporators for zeotropic refrigerant mixtures. Int. J. Refrig., Vol. 19., No. 6, pp. 361-368, 1996.
Venkatarathnam, G., Murthy, S. S. Effect of mixture composition on the formation o f pinch-

points in condensers and evaporators for zeotropic refrigerant mixtures. Int. J. Refrig., Vol. 22.,
No. 3, pp. 205-215, 1999.
Wang, S-P., Chato, J. C. Review o f recent research on heat transfer with mixtures - Part 2:
boiling and evaporation. Proc. ASHRAE Transactions Symposia, paper no. CH-95-23-3, pp.
1387-1401, American Soc. Heating, Refrig. Air-Condition. Eng., Atlanta, USA, 1995.
241
APPENDICES
Appendix 1: Information on refrigerant mixtures

Table 31 is a list of commercial refrigerant mixtures taken from UNEP (2002), including their compositions as mass fractions. Saturation pressure (at 0°C dew-point
temperature), the corresponding temperature glide, and mean glide shape index ( nG) were calculated from Refprop (Lemmon et al, 2002).
Table 31: Commercial zeotropic mixtures and selected properties
Component Component mass fraction P at TdP = APGat Tdp = Glide shape

Refrigerant
1 2 3 4 1 2 3 4 0°C (kPa) 0°C (K) index, nG (-)
R-401A R22 R152a R124 0.53 0.13 0.34 343 6.1 0.95
R-401B R22 R152a R124 0.61 0.11 0.28 368 6.0 0.88
R-401C R22 R152a R124 0.33 0.15 0.52 280 5.6 1.14
R-402A R125 R290 R22 0.60 0.02 0.38 660 0.8 1.17
R-402B R125 R290 R22 0.38 0.02 0.60 621 1.3 1.07
R-403A R290 R22 R218 0.05 0.75 0.20 643 2.2 0.81
R-403B R290 R22 R218 0.05 0.56 0.39 669 0.1 0.48
R-404A R125 R143a R134a 0.44 0.52 0.04 613 0.5 0.70
R-406A R22 R600a R142b 0.55 0.04 0.41 336 9.1 0.94
R-407A R32 R125 R134a 0.20 0.40 0.40 559 5.3 0.93
R-407B R32 R125 R134a 0.10 0.70 0.20 617 3.1 0.78
242
R-407C R32 R125 R134a 0.23 0.25 0.52 516 5.8 1.05
R-407D R32 R125 R134a 0.15 0.15 0.70 432 5.0 1.27
R-407E R32 R125 R134a 0.25 0.15 0.60 485 5.9 1.14
R-408A R125 R143a R22 0.07 0.46 0.47 569 0.4 0.99
R-409A R22 R124 R142b 0.60 0.25 0.15 353 8.3 0.89
R-409B R22 R124 R142b 0.65 0.25 0.10 372 7.8 0.83
R-410A R32 R125 0.50 0.50 775 0.3 0.92
R-410B R32 R125 0.45 0.55 767 0.4 0.94
R-411A R1270 R22 R152a 0.02 0.88 0.11 471 1.8 0.76
R-411B R1270 R22 R152a 0.03 0.94 0.03 498 0.8 0.78
R1270 R22 R152a 0.03 0.96 0.02 503 0.5 0.85
R-412A R22 R218 R142b 0.70 0.05 0.25 439 9.9 0.86
R-413A R218 R134a R600a 0.09 0.88 0.03 341 2.0 1.30
R-414A R22 R124 R600a R142b 0.51 0.29 0.04 0.17 330 8.3 0.97
R-4I4B R22 R124 R600a R142b 0.50 0.39 0.02 0.10 324 8.1 1.00
R-415A R22 R152a 0.82 0.18 445 2.3 0.80
R-415B R22 R152a 0.25 0.75 304 1.3 1.26
R22 R152a 0.52 0.48 364 2.5 1.09
R22 R152a 0.60 0.40 384 2.7 1.02
R-416A R134a R124 R600 0.59 0.40 0.02 254 2.0 0.87
R-417A R125 R134a R600 0.47 0.50 0.03 480 4.1 0.98
243
R-418A R290 R22 R152a 0.02 0.96 0.03 500 0.8 0.94
R22 R124 R600 0.50 0.47 0.03 325 7.6 1.02
R23 R32 R134a 0.05 0.22 0.74 498 9.3 1.82
R23 R125 R143a 0.20 0.36 0.44 997 9.1 1.53
R32 R125 R134a R600 0.10 0.42 0.45 0.03 518 4.8 0.97
R32 R125 R143a 0.10 0.45 0.45 666 0.5 1.18
R32 R125 R143a R134a 0.02 0.41 0.50 0.07 610 1.1 0.77
R32 R125 R143a R134a 0.10 0.33 0.36 0.21 591 3.2 0.85
R32 R134a 0.25 0.75 421 5.1 1.34
R32 R134a 0.30 0.70 447 5.6 1.26
R125 R22 0.70 0.30 654 0.2 0.84
R125 R134a R152a 0.35 0.40 0.25 426 5.2 1.12
R125 R143a R290 R22 0.42 0.06 0.02 0.50 631 1.1 1.11
R125 R152a R227ca 0.40 0.05 0.55 418 9.9 1.03
R134a R142b 0.80 0.20 268 1.5 0.73
R134a R142b 0.81 0.19 269 1.4 0.73
R134a R142b 0.88 0.12 278 0.9 0.72
R170 R290 0.06 0.94 570 4.3 2.13
R290 R600a 0.50 0.50 304 7.9 1.06
R600a R600 0.50 0.50 130 1.1 1.01
244
Appendix 2: Calculation of HTF properties
Water was chosen arbitrarily as the HTF used in the heat exchanger models. Property data from
Incroprea and De Witt (1990) was fitted to b^-order pohnomial equations to enable their rapid
calculation at selected temperatures within the code (equation 264).
{ p , p , Cp, k ) = c } -T6 + c 2 -t 5 + c 3 -t 4 + c4 -t * + c 5 -t 2 + c 6 -t + C j (264)
where { p, p, Cp. k } refers to the selected property, and c l .. ,c7 arc specific to each property- and
are listed in Table 32. r is a function o f the absolute temperature according to equation (265).
t = ------ — ------ (265)

273.15 + 7 ^
where Thtf is the local HTF temperature.
Comparison o f equation (264) with the original data showed an error of less than ±0.1%.
Table 32: Constants fo r water property polynomial equation (265)
Property { p . p , Cp. k }
Constant Specific heat Thermal
Density ( p ) Viscosity ( p )
( Cp ) conductivity ( k )
Ci 415.370 -0.104601 -0.570263 -0.003163
c, -7901.99 2.02702 10.79505 0.060920
62462.2 - 16.0464 - 84.95% -0.486447
-262581 66.4839 355.880 2.06264
C5 619076 - 151.948 -836.950 -4.90033
C6 -775931 181.150 1047.92 6.18660
C1 404721 - 83.349 - 545.067 -3.24324
245
Appendix 3: Overview of new balancing technique
Table 33 gives a summary of the stages required for the new balancing technique. The second
equation listed in stages 1, 2 and 3 refer to the proposed numerical form o f the characteristic
expressions
Table 33: Overview o f new balancing technique
Stage Procedure Equation

1. Use compressor characteristic data for: (251.
Compressor 191)
data Q e .c o m p ~ f i f e . d p i o t B i ' T . d p i m ) )
(252.
Q cx o ip T \ f t , d p i o ta 1' ,d p i im ) ) 192)
■ ^ d sk ~ /{T .d /H w O 'T .d /H m ) ) (250.
194)
2. Use condenser characteristic data for: (249,
Condenser 217/218)
data Q c .c o n d = f ( 1' )
(256,
T .d p i o u t ) = f { T . d p { i * ) ? ^ T i h ) 215)
3. Evap Use evaporator characteristic data for: (255.
data 231/232)
Q e .e v a p ~ f (T'e , d p ^ o t t t \ e ,ip (o u i) )
4. Taking the relevant compressor characteristic data (equation 252) and the
Condenser relevant condenser characteristic data (equation 249). new condenser
charac. characteristics are determined by matching the condenser capacity to
matched to desuperheat. This provides a condenser characteristic that is a function of
compressor both condensing temperature and evaporating temperature:
desuperheat
Q c .c o n d ~~ f c . d p ( i n ) ' ^ ' e . t f y [ , < m l ) )
(253)
5. Applying the new condenser characteristic (equation 253) to the
Condensing compressor characteristic (equation 252) provides evaporating temperature
unit charac. equilibrium points which, when linked, provide a single condensing unit
characteristic:
Q c .c u (T c .d p im ) ) = / ( T e ,d p (o u l) ) (254)
6. Using the compressor desuperheat characteristic (equation 250) with the
Evaporator condenser Ap conversion characteristic (equation 256) the corresponding
charac. value of tcdpioun is found for Tcdp{in) for the condenser and compressor.
matched to
condenser Using the evaporator characteristic (equation 255) a new set of evaporator
inlet data is obtained from selecting an evaporating capacity based on
evaporating temperature and the value of Tcdp[m) that corresponding to the
Tc .c ty o in ) :
Q e e va p ~ f ^ e .d p io u t) ) (257)
246
Stage Procedure Equation
7. Applying the new evaporator characteristic (equation 257) to the
Evaporatin compressor characteristic (equation 251) provides condensing temperature
g unit Tcdpim) equilibrium points which, when linked, provide a single evaporating
charac. unit characteristic:
=f { T edp{out) )
(258)
8. Convert Using the compressor characteristic (equation 252) and the condensing unit
unit characteristic (equation 254), convert the condensing unit capacity from
capacity condensing capacity to ev aporating capacity:
charac.
Q t . c /u ^ e .d p io u t) ) = f \ j e d p i i n ) )' 0F
(258)
Using the compressor characteristic (equation 251) and the ev aporating
unit characteristic (258) convert the capacity7variable from evaporating
capacity to condensing capacity7:
Q t.e = f { ^ , .d p ( o u l ) ) (259)
9. Balance Apply the condensing unit evaporating capacity (equation 258) to the
condensing ev aporating unit evaporating capacity (equation 259) to obtain the system
unit with equilibrium point:
evaporating
Qt t ,M(Tcdp{ai)) or
unit
Jf [VTt.d p ( 0*1) J " Jf [\Te .d p io v t) V
/ (260)
Apply the condensing unit condensing capacity (equation 254) to the

evaporating unit condensing capacity (equation 259) to obtain the system
equilibrium point:
Q c .c l u ^ T e .d p d m i ) ) Q c .t i u ( ^ e .d p { o t a ) ) _q
/(^.4>(ui) ) f{ ^ c .d p {m ) )
(261)
247
Appendix 4: Results of worked examples for balancing
Table 34 present the main results from the different balancing exercises referred to in Section 5.
Table 34: Comparison o f equilibrium-points fo r example balancing exercises

Evaporator Condenser
Example Balance
system technique T e.dP Qe Error T c,dP A T dsh Q c Error
(°C) (kW) (%) (°C) (X) (kW) (%)
model -10.46 3.26 47.40 35.4 5.33
System 1 (a); new -10.46 3.31 1.5 47.40 5.28 -0.9
small
compressor mid-pt -9.74 3.47 6.4 47.40 - 5.45 2.3
dew-pt -8.54 3.63 11.4 47.98 - 5.66 6.2
model -13.73 2.91 44.57 38.8 4.76
System 2 (a); new -13.73 2.98 2.4 44.57 4.79 0.6
small
dew-pt -13.03 3.09 6.2 44.62 - 4.93 3.6
model -10.19 3.45 46.25 35.0 5.37
System 3 (a); new -10.19 3.41 -1.2 46.25 5.36 -0.2
small
dew-pt -7.28 3.87 12.2 47.54 - 5.93 10.4
model -13.06 3.10 43.66 37.9 4.89
System 4 (a); new -13.06 3.12 0.7 43.66 4.94 1.0
small
dew-pt -12.28 3.25 4.8 43.79 - 5.10 4.3
model -18.93 4.15 56.58 46.1 8.49
System 1 (b): new -18.93 4.33 4.3 56.58 8.47 -0.2
large
dew-pt -16.63 4.81 15.9 57.43 - 9.21 8.5
model -21.62 3.94 53.49 50.3 7.76
System 2 (b); new -21.62 3.97 0.8 53.49 7.80 0.5
large
dew-pt -20.18 4.31 9.4 53.48 - 8.29 6.8
model -18.91 4.34 55.73 46.0 8.53
System 3 (b); new -18.91 4.40 1.4 55.73 8.53 0.0
large
dew-pt -16.13 4.99 15.0 57.03 - 9.42 10.4
model -21.18 4.07 52.90 49.5 7.93
System 4 (b); new -21.18 4.11 1.0 52.90 7.97 0.5
large
dew-pt -19.61 4.49 10.3 53.02 - 8.51 7.3
248
PART 3: APPRAISING TH E FLA M M A BILITY HAZARDS OF HYDROCARBON
R EFRIG ER A N TS
249
NO M ENCLATU RE
Letters
a a constant
A area (m2)
area o f air discharge vent (m2)
A«r effective area (m2)
Af flammable area (m2)
Ar area o f room floor (m2)
^R m area o f room floor (m2)
A vent area (m2)
vent panel area (m2)
^ wail wall area (m2)
b a constant
Br Bradley number (-)
c concentration (kg m'3)
c mean concentration (kg m 3)
actm actual measured concentration (kg m'3)
Cd discharge coefficient (-)
Cd mean concentration with airflow (kg m'3)
c f
floor concentration (kg m°)
Cf mean floor concentration (kg m 3)
C F flammable concentration (kg m'3)
f.calc calculated mean floor concentration (kg m'3)
c J.. max maximum floor concentration (kg m 3)
r f.m ax maximum mean floor concentration (kg m 3)
r max.des maximum design concentration (kg m'3)
c ,.d mean concentration without airflow (kg m‘3)
c Rm homogenous room concentration (kg m'3)
c surr concentration of surroundings (kg m°)
wind wind pressure coefficient (-)
D diameter (m)
E an event (-)
f frequency ( y 1)
r frequency o f ignition ( y 1)
f fa ta l fatality frequency ( y 1)
f leak leak frequency ( y 1)
g gravitational acceleration (m s'2)
g' reduced gravity (m s'2)
251
h height (m)
H enthalpy (J kg'1)
hd height o f air discharge (m)
hgff effective height (m)
H mot molar enthalpy (J kmol'1)
hr height o f release (m)
hRm room height (m)
hthe™ height of thermal source (m)
1 thermal intensity (s (kW m'2)40)

I^ thermal intensity to room occupants (s (kW m'2)43)
K deflagration coefficient (Pa m s'1)

LF1. lower flammability limit (kg m'3)
m mass (kg)
m fiam mass o f burned mixture/combustion products (kg)
mF flammable mass (kg s 1)
mF mean flammable mass (of refrigerant) (kg s'1)
m Fa mean flammable mass o f refrigerant and air mixture (kg s'1)
mr mass o f refrigerant (kg)
mr mass flow rate of release (kg s'1)
mr a mass of refrigerant and air (kg)
mr e entrained mass of refrigerant (kg)
mrmax max allowable mass of refrigerant (kg)
m r Rm mass of refrigerant within room (kg)
mRm,tot total mass of refrigerant and air within room (kg)
mv vented/ventilated mass (kg)
252
n an index (-)
N number o f items (-)
N conn number o f connections (-)
number o f spatial elements in descending plume (-)
X* number o f room radial elements (-)
Nf number o f events in reference time (-)
N leak number of leak sizes (-)

mol
number o f moles (-)
H occ number o f occupants (-)
X so, number o f sources o f ignition (-)
N S O ,J number o f sources o f ignition at floor level (-)
n t number of outside temperature increments (-)
^ wind number of wind speed increments (-)
253
p pressure (Pa)
p probability (-)
P a tm atmospheric pressure (Pa)

Pavail probability o f availability (-)
P D potential difference (V)

Pe Peclet number (-)
P F* probability o f ignition (-)
P fa il probability o f mechanical failure (-)
P probability o f fatality (-)

fatal
A P /h friction pressure difference (Pa)
pL probability of source o f ignition (-)
p>
f probability of flammable volume (-)
p leak probability (-)
leak
Plocal probability o f local leakage (-)
P max maximum pressure (Pa)
Ap° overpressure (Pa)
&P°Rm room overpressure (Pa)
*P °P
overpressure necessary to remove vent panel (Pa)
P
perc percolation probability (-)
P s ta static pressure (Pa)
^ P stack stack pressure difference (Pa)
probability o f leak from a sy stem (-)
A P to t total pressure difference (Pa)
Pvel velocity pressure (Pa)
P v p ,fix vent panel fixing pressure (Pa)
A P wind wind pressure difference (Pa)
q heat flux (W m'2)
0 thermal load, output (kW)
q occ heat flux radiated to occupants (kW m'2)
254
r radius (m)
Ri Richardson number (-)
local Richardson number for cloud/spreading plume (-)
S a set o f events (-)

t time (s)
T temperature (°C)
t' reference time (s)
Ta air temperature (°C)
t* post-ventilation period (s)
^d,rmn min post-ventilation period (s)
event interval time (s)
t fad time until repair of failure (s)
tF flammable time (s)
^flam combustion/burning time (s)
T
1 flam flame temperature (°C)
^ leak leak time (s)
^leak,cat catastrophic leak time (s)
^mech operating time of mechanical ventilation (s)
ôcc occupancy time (s)
f pl
plume descent time (s)
*r release time (s)
tres residence time (s)
1Rm room temperature (°C)
^sprd floor spreading time (s)
^therm presence time o f thermal sources (s)
^tot total time (s)
ûnit operating time o f RAC unit airflow (s)
255
u velocity (m s'1)
u fii
friction velocity (m s'1)
UFL upper flammability limit (kg m°)
U flam burning velocity (m s'1)
^ flam,lam laminar flame velocity (m s'1)
^ f,s p r d spread velocity across floor (m s'1)
UP
plume velocity (m s'1)
U Rm mean room air speed (m s'1)
U ss sonic velocity (m s*1)
U uind wind speed (m s'1)
V volume (m3)
V volume flow rate (m3 s'1)
K air volume flow rate (m’ s'1)
Vd discharge air volume flow rate (m3 s'1)
K entrained volume (m3)
vF flammable volume (m3)
VF mean flammable volume (mJ)

vflam volume of burned mixture (m3)
K n fli infiltration volume airflow rate (irf s'1)
vtnech mechanical ventilation volume airflow rate (m3 s'1)
plume volume (m3)
v, plume volume flow?rate (m3 s'1)
K refrigerant volume flow rate (m3 s*1)
v*. room volume (m3)
vtherm volume airflow rate from thermal sources (m3 s'1)
V umt unit volume airflow rate (m3 s'1)
K vented/ventilated volume (m3)
Wvp vent panel plane density' (kg m'2)
256
X mass fraction (-)
X flam mass fraction of burned mixture/combustion products (-)
* r . mass fraction of flammable refrigerant/air mixture (-)
mass fraction of refrigerant/air mixture (-)
z distance (m)
^ flam burning distance (m)
^ occ distance to occupant (m)
Greek
a entrainment coefficient (-)
a c local layer entrainment coefficient for cloud/spreading plume (-)
a P
plume entrainment coefficient (-)
P mixing exchange ratio (-)
X 1OCC integrated thermal intensity-frequency (s (W rfr2)4'3 y'1)
X Apo integrated overpressure-freq. (Pa y'!)
X integrated risk factor (-)

£ mixing effectiveness (-)
E expansion ratio (-)
* time fraction (-)
$ in fll time fraction o f infiltration rate (-)
$ mech time fraction o f mechanical ventilation (-)
$ OCC occupancy time fraction (-)
$ op time fraction o f on-cycle operation (-)
$ therm time fraction of thermal source (-)
$ u nit time fraction of unit airflow (-)
r fuel radiative fraction (-)
<p safety factor (-)
257
X friction factor (-)
A room air changes ( h 1)
A 50 reference infiltration air changes at 50 Pa ( h 1)
infiltration air changes ( h 1)
mech mechanical ventilation air changes ( h 1)
ûml unit airflow air changes ( h 1)
M molar mass (kg kmol-1)
P viscosity (Pa s)
n pi-group for dimensional analysis

0 air discharge angle (rad)
p density (kg m'3)
Pa density of air in room (kg m°)
P flam density of burned mixture/combustion products (kg m'3)
Pr density o f gas at room conditions (kg m°)
Pv density of mixture at vent (kg m°)
T turbulence factor (-)
Tfn fiction sheer stress (N m)
s incidence ratio (-)
m deflagration outflow interaction parameter (-)

CO mass fraction of element (-)
V refrigerant retention ratio (-)
S angle adjustment factor (-)
258
Subscripts
0 condition at origin
h height level element
i environmental condition
in into element
j dispersion temporal/spatial element
k airflow source
I leak size element
leak leakage
max maximum value
min minimum value
out out o f element
r refrigerant
Superscripts
* ignition
F flammable
mol molar
SECTION 1: INTRODUCTION
1.1 Hydrocarbon refrigerants

Hydrocarbons (HCs) are increasingly being used as alternatives to fluorinated refrigerants such
as HCFCs and HFCs because of environmental issues.17 However, as they are flammable, this
raises a safety issue not normally associated with conventional refrigerants. A substantial release
of flammable refrigerant into a room that has not been specifically designed to accommodate
flammable concentrations - such as an office environment - could result in fire and/or explosion.
Safety standards are written as a step towards achieving an acceptable safety level when using
HCs, but they should be used in conjunction with a risk assessment to further help mitigate the
associated hazards.
1.1.1 Safety standards and codes

There are safety standards that provide guidance on the use and application o f refrigerants, and
these include specific requirements for flammable refrigerants. However, they do not necessarily
provide insight into risk assessment methodology, refrigerant leakage, dispersion and
consequences of ignition; instead, they offer practical guidance on best industry practice.
Within Europe, the generic standard for safety of refrigerating systems is EN 378: 2000. This
classifies refrigerants according to their flammability and toxicity, and system design and
installation requirements are given for each refrigerant classification. Flammable refrigerants are
classified according to certain properties: a fluid with a low Lower Flammability Limit (LFL) and
a large heat of combustion (HOC) is designated “A3” and those that have higher LFLs and small
HOCs are “A2”. R290 and R600a are examples o f A3, and R152a and R32 are examples o f A2.
The standard specifies the maximum quantity of refrigerant permitted in a location, whether to
employ ventilation, refrigerant detection and/or other safety features, and details requirements for
various system components. Equipment using flammable refrigerants must be devoid o f sources
of ignition. For systems located within occupied spaces, the maximum refrigerant mass per
circuit is specified, for example, 2.5 kg o f an A2, and 1.5 kg of an A3 refrigerant in publicly
occupied spaces. However, to avoid relatively large quantities o f refrigerant being released into
small spaces, an additional limit is specified in terms of a maximum mean concentration. This is
termed the “practical limit” and corresponds to 20% of the LFL, being approximately 8 g m 3 for
HCs. For example, a room of 100 m3, the maximum refrigerant charge within a single refrigerant
17 This is addressed in detail in Part 1.

260
circuit would be 0.8 kg, whereas a system in a 250 m 3public occupied room would have its
charge limited to 1.5 kg (rather than 2.0 kg).
Other than the generic standard, there are specific product standards such as EN 60335-2-24 and
EN 60335-2-89, which cover domestic and commercial refrigerators respectively. Both specify a
maximum charge o f 150 g o f any flammable refrigerant regardless of flammability classification,
room size and occupancy type; if this amount is exceeded, then EN 378 must be used. Also,
industry codes such as ACRIB (2000) and IOR (2000) contain practical guidance on safety
measures.
1.1.2 Risk assessment for refrigerating systems

The concept of “risk assessment'’ can be applied qualitatively (i.e., “what if?”), or quantitatively'
(i.e., determining the probability/frequency o f the “what i f ’). In either case, the first stage is to
conduct a qualitative assessment, so that potential hazards are identified, which in the case of a
flammable refrigerant is ignition o f a release. For this to occur, three criteria must be satisfied: a
release of refrigerant, mixing of refrigerant and air to within the flammable range, and the
presence o f an “active” source o f ignition (SOI).
For a release o f flammable refrigerant into air, locally there will always be a gradient from 100%
refrigerant to pure air, so it follows that a flammable concentration will always occur. An active
SOI is one that has sufficient energy to initiate combustion between the flammable material and
the oxygen in the air. This energy' may be in the form of heat (e.g., from an electric heater), a
spark (e.g., from an arc between electrical contacts) or a chemical reaction. Thus, the presence of
a flammable mixture at the same time and location as the active SOI may result in ignition.
Minimisation o f this occurrence is necessary to reduce the risk.
Consideration o f the construction, installation and operation of typical refrigeration and air
conditioning (RAC) equipment provides an indication o f the reduction measures available. A
leak-tight refrigerant circuit contains refrigerant and is therefore prevented from mixing with air
and reaching SOI. Containment can be achieved in most situations through proper design and
installation, although unforeseen, external and human behavioural influences can lead to leakage.
Leakage issues are well documented but whilst containment should be a priority during design, it
cannot be employed as the sole mitigation technique to prevent ignition.
261
Whilst the potential SOI associated with equipment can be eliminated, the SOI within an
occupied space cannot be controlled, so the only option is to minimise the size (or volume) and
duration o f a flammable concentration; this concept is core to the study.

This literature review draws on studies relating to two main areas: risk assessment methodolog>:
for flammable gases, and more specifically, the behaviour o f a release of flammable gas within a
confined space.
1.2.1 Risk assessment of flammable gas

Given the common use o f flammable materials, literature on appropriate quantitative risk
assessment (QRA) methodology is extensive, although only a selection of the most relevant are
discussed presently. A set of publications that provide comprehensive information on the subject
were published by the Dutch Government. The ‘“Purple Book*’ (CPD, 1999) details the general
approach and formulae for risk calculations, and calls upon the “Red Book” (CPD, 1988) for
calculating probabilities, the “Yellow7Book” (CPD, 1997) for methods and data on release and
dispersion, and the “Green Book” (CPD, 1992) for methods and data on consequences of ignition.
Due to widespread use of HCs as fuel gas (e.g., liquefied petroleum gas, LPG), QRA is
frequently used and various models have been developed. Khan and Abbasi (1999 and 2000)
present a QRA method (as used for their commercial software) for addressing the release of HCs
in refineries. It was designed for major discharges of flammable material, and employs a number
o f models selected according to the type and location of the release. These deal with dispersion of
an outdoor cloud and subsequent fire and/or explosion. The model outputs frequencies and
severity7of consequences. Ditali et al (2000) describe a similar approach (also as commercial
software) but focused on LPG storage vessels. Both methods use a simple steady flow discharge
equation and liquid spread model for the initial release, and an outdoor dispersion model to
determine flammable concentrations. The consequence is limited to radiation emitted from the
resulting fire. Some results are given, but no empirical data on aspects such as leakage and
ignition sources was given. An industrial QRA model developed by Barry (1992) addressed
failure probability of components within a LPG plant, and also presents an approach for
calculating ignition frequency at time steps from initiation o f the release.
Whilst these studies focus on outdoor releases, the only study found on indoor HC releases was
Lee (2002). Although it mainly deals with consequences of toxicity7in the case o f a spill the
262
methodology was still relevant since calculations addressed discharge, evaporation and dispersion
o f the release, and estimation of fatality from a given concentration; thus analogous to fatality
from explosion effects. Rew et al (2000) evaluated ignition probability models for LPG releases
from off-site storage vessels, and compared the results with a newly developed model.
Calculations o f the existing and new methods relied on various coefficients dependent on location
type (urban, rural and industrial, indoor/outside) and time of dav/night. Ignition probability' was
estimated for a range o f locations and size of flammable clouds.
1.2.2 Risk assessment of flammable refrigerants

To date, many QRAs that address the use o f flammable refrigerants in different types of
stationary' RAC equipment have been published, as well as some for vehicle air conditioners and
transport refrigeration. The latter were not considered relevant and therefore not discussed.
Several QRAs were based on the approaches detailed in the Purple Book (CPD, 1999). Adopting
its principles, Blom-Bruggeman et al (1996a) published a generic method specific to the context
o f RAC equipment, covering both flammable refrigerants (such as HCs) and ammonia.
Subsequent studies by Blom-Bruggeman et al (1996b) and Elbers and Verwoerd (1997)
addressed risk o f fatality' from HCs used in bulk milk tanks and domestic heat pumps
respectively. Normal operation, service/maintenance and charging activities were included in
both, and the methodology' employed was based on a fixed set o f installation characteristics and
environmental conditions. Evaluation of leak frequencies, release size, dispersion o f the release
and fire/explosion consequences was made with separate commercially available models designed
for industrial process plant. Values used for ignition probabilities were constant and releases
inside rooms w'ere assumed to disperse evenly, w hilst more sophisticated models were employed
for releases outside. A more recent study from the same group (van Blanken and Verwoerd,
2001) assessed a commercial freezer cabinet in a number o f different locations. The approach
was similar to those detailed above, although results of ignition tests using cigarettes were
included, and an explosion from within the compressor compartment was modelled. Publications
including van Gerwen and Jansen (1994), van Gerwen and Koffijberg (1995) and van Gerwen
and Verwoerd (1996) reiterate the results o f these studies.
A series o f QRAs for a range of commercial refrigeration equipment, including supermarket chest
freezers, integral supermarket display cases, and drinks chillers in public houses were produced
by ADL (1998). These were based on detailed fault tree analysis specific to the equipment and
installation, assuming three activity modes (normal operation, defrost and maintenance). They
263
called on separate release size, dispersion and fire/explosion software, and leak frequencies were
from consultation with manufacturers and contractors. Dispersion modelling used a simplistic
approach based on one fixed set o f environmental conditions, and SOI were specified according
to the application.
One o f the earliest works relating to QRA of flammable refrigerants is Nowacki (1995), w hich
summarised a basic methodology. Software was provided, which calculates societal risk (i.e., risk
to the local population) as a function o f refrigerant mass and characteristics of the surrounding
area (such as population density). Historical data for LPG incidents was used to estimate the
number of injuries and fatalities associated with an installation of a given mass o f flammable
material. Thus, the approach is basic and takes no account o f specific characteristics o f the
equipment or the environmental conditions.
Wolfer and Seiler (1999) and Wolfer (1999) report on a QRA for three different applications: a
domestic heat pump, an indirect central supermarket system and an integral display cabinet.
Probability calculations were made according to Blom-Bruggeman et al (1996a) and a pre
published draft of EN TR 14739 (2004). Whilst the report contains lengthy discussion on the
characteristics o f each application, there was minimal input data: descriptions o f the calculation
procedures were brief and the dispersion was treated simplistically. One interesting aspect of the
study w as a comparison of risk associated with the heat pump versus that of a gas heating
system.
Recently, the standards committee responsible for EN 378 has published a report (EN TR 14739:
2004) on the methodology for conducting a QRA on domestic refrigerators. The methodology is
simplistic in that basic empirical data are given for use in basic calculation procedures. However,
unlike most other references it considers stages other than in-use, such as transportation of
appliances, servicing and disposal. For each stage comprehensive flow charts are given. Given
that the purpose o f the document is methodology, it contains to no results.
A comprehensive study by Goetzler et al (1998) addressed flammable fluorocarbon refrigerants

in residential split air conditioners. Part of the study reported on extensive ignition test conducted
on common devices to determine which were potential SOI, and a number of leak simulation tests
to characterise refrigerant concentrations. Conventional fault tree analysis was used to estimate
frequency of fires in different household locations.
264
Overall, these studies cover a wide range o f RAC equipment and charge sizes and most address
operation and servicing activities separately. Results comprised different forms o f risk, including
occurrence of fire, explosion, secondary fire, injury, fire damage, although the majority use
frequency o f fatality. Frequencies for fire during normal operation vary considerably from 10'9 to
1CT4y '1, whereas fatality is typically 10*6 to 10'8 y'1. It was not possible to make comparisons
between the results because each was based on unique equipment and applications. Specific
details of each are provided in a summary in Appendix 1 (Table 55).
1.2.3 General gas dispersion in confined spaces

The subject of gas dispersion is vast, and therefore it w as necessary to establish the main
characteristics relevant to this study, and useful information was found in Britter (1988), CCPS
(1996) and Woodward (1998). Also, the mixing process is complex and dependent upon a
number of factors, so know ledge of release characteristics and surrounding environment is
essential. First, releases can be split into two categories: releases to the open atmosphere (outside)
and released within confined spaces (inside). O f relevance to this study are releases within
confined spaces, so most literature related to outdoor dispersion was not covered. Given that all
refrigerants of interest have a density greater than air, studies on light and neutrally buoyant
releases were also neglected. Further, releases are also differentiated in terms o f instantaneous (or
puff), and continuous releases.
A leak initially produces a jet or plume for w hich “aspect ratio'’ - or the size o f the plume to the
room dimensions - is an important factor. The difference between low' and high aspect ratios are
characterised by dominant mixing in the descent of a plume as opposed to mixing occurring only
at ground level in the case of a low aspect ratio release. There are two basic mechanisms by
which mixing occur. One is diffusion w hich is caused by random motion o f particles across a
concentration differential. The second mechanism is convective mixing, comprising buoyancy and
externally driven mixing.
Most experiments have shown the rate of convective mixing, caused by density' difference and/or
external air movement, to be proportional to a (vertical) “stratification parameter” (Britter,
1988), termed the Richardson number ( R i ). This is used to characterise the work required to mix
a stable stratified layer and it depends upon the environment and the nature of the release such as
the buoyancy or externally induced convection. Normally both diffusion and convection are
present and a measure of their relative impacts is the Peclet number (Pe ). When Pe / Ri > 6, the
effect of diffusion is negligible (Britter, 1988). Another significant parameter (particularly for
265
low aspect ratio releases) is the texture o f the surface below the cloud. A rough surface causes
drag so the horizontal or spreading velocity decreases, which inhibits mixing.
Some investigations considered the influence o f room objects on dispersion. Experiments by

Leach and Bloomfield (1973) indicated that objects such as furniture had a negligible effect on
mixing. Britter (1988) confirmed that experiments with room obstacles generally show that they
divert the direction o f the gas flow and in doing so increase turbulence, leading to a higher rate of
mixing. The exception is where structures channel the flow thereby reducing the exposed area and
thus mixing rate. A small number o f tests by Clodic (1997) used a 100 mm barrier laid across the
room floor. Contrary to the conclusions of others, it was found that floor concentrations w ere
higher than that without the barrier, comparable to a release into a room with a smaller floor
area.
The influence of household furniture w as further investigated by Jabbour and Clodic (2002) and
Jabbour et al (2003). Tests demonstrated refrigerant accumulated beneath a sofa, although
elsewhere the refrigerant dispersed as would be expected. Also noted was the influence of a thick
carpet which was found to “accumulate'’ refrigerant whilst better mixing occurred elsewhere. The
carpet increased the surface roughness, which corresponds to observations by Britter (1988).
1.2.4 Releases into quiescent rooms

There were a number of studies relating specifically to releases into “still'’ enclosures, absent o f
forced airflow. An early study by Baines and Turner (1969) addressed the development of a
single plume by entrainment in an enclosed space of high aspect ratio. A generalised model w as
described that accounts for density difference, release velocity' and diameter, shown to correlate
closely with measurements. An illustration o f the dispersion process is shown in Figure 95(a),
where the plume impinges on the floor surface to produce overturning or continuous pushing up
o f the preceding layer, creating stratified layers. As the layer depth increases, the mixture is “re
entrained” into the plume as it passes through the layer. Whilst the observations o f Baines and
Turner apply to a downw ard release of a dense refrigerant, a paper by Baines et al (1990)
provided insight into the situation where a release is pointed upward thereby forming a
“fountain”. By comparison with the downward release (as Figure 95a) considerable turbulence
developed when the plume reversed its direction with the consequence o f more effective mixing.
Various authors reported on continuous releases into (full-scale) rooms. Fletcher and Johnson
(1986) continuously released C 0 2 into an unventilated room, and their results mirrored those of
266
Baines and Turner (1969). At any time during the release, concentration at the floor was highest,
and gradually reduced with height up to the leak point, with the concentration profiles represented
by a parabolic relation with height (dotted curve in Figure 95b). A release from low level
produced a much higher floor concentration since the distance for the plume to descend was
insufficient to establish sufficient buoyancy driven mixing and entrainment. A release of low er
density gas tended to produce lower concentrations and to an extent that a homogenous
concentration occurred when the release point was above three-quarters of the room height.
More detailed work by Marshall (1983) showed the profile remained constant over time (when
forced airflow was absent) whilst continued gas flow caused the overall concentration within the
region to increase; this is consistent with Britter's work (1998). Following Fletcher and Johnson
(1986), Marshall also demonstrated concentrations w ere sensitive to the direction o f the leak
especially where the release was close to the stratified layer. With a release from the ceiling, the
release direction had a lesser influence on floor concentration. Positioning of leak source at
different lateral locations within the room had negligible effect. Further experiments by Marshall
indicated that releases into single rooms were representative of releases into complex structures
such as two or more rooms linked by an open passage, or rooms with irregular geometries.
release source
upper layer
plume
concentration
Figure 95 (a) Layer formation in a quiescent room and, (b) concentration profiles
The concept o f Baines and Turner (1969) was adapted by Cleaver et al (1994) - in the form of a
non-dimensional correlation - to describe the behaviour o f gas build-up in quiescent rooms
according to release characteristics and room dimensions. From experiments with positively and
negatively buoyant gases, it w as observed that jet length, momentum and buoyancy determined
the mixing rate, and formation of a stratified layer. In line with Marshall's work, measurements
267
revealed two t>pes of gas layer, as seen in Figure 95(b): one of homogenous concentration and
another where concentration reduced from that of the first layer to zero (at the height o f the
source). Additionally, high velocity releases tended to mix well, leading to a near-homogenous
room concentration.
Rabkinkov (1988) presented concentration measurements for a number o f continuous releases

into a room from different heights over several time increments. The results were consistent with
Cleaver et al (1994), and Fletcher and Johnson (1986). In contrast to the approach of Cleaver et
al, a diffusion theory -based model was developed, w hich show ed good agreement with the
measurements.
In a study specific to releases of flammable refrigerants, Kataoka et al (2000a, 2000b) considered

the case of a rapid release from an air-conditioner. The dependency of the development of floor-
level concentration on several variables was evaluated for a high aspect ratio situation, including
release height, velocity and floor area. Again, velocity was shown to have a significant effect on
mixing rate (especially at low level), and floor concentrations reduced as leak height and gas
density increased, which results generally agreed with Cleaver et al (1994), Marshall (1984) and
Fletcher and Johnson (1986). Room tightness was generally found to affect floor concentrations,
since even a room with minimum fabric leakage produced lower concentrations than a completely
sealed room. Using measurements, the concentration-height profile was characterised with a
parabolic equation, so floor concentration could be estimated if floor area and release height and
mass were known.
A series o f refrigerant leak simulation tests w ere reported by Clodic and Cai (1996, 1997) and
Clodic (1997). R290 and R600a were released through a capillary tube from freezers and
portable air-conditioners, over a range of leak rates, heights and positions within the room. A
clear relationship was found between release height and maximum floor concentration, although
in contrast to some other studies, the results did not show an obvious correlation betw een
increasing release flow rate and floor concentration for either refrigerant. Releases from non-
central lateral positions (particularly at room comers) produced higher floor concentrations.
Following cessation of the release the denser R600a was found to reside as a stratified layer for
longer periods than R290.
Zhao et al (2002) addressed the dispersion o f a jet - independent o f the room and airflow
conditions - and presented a model to characterise local concentrations. The steady flow7equation
268
for sonic flow was used to estimate leak exit velocity from different parts of a refrigeration
circuit. The concentration field was calculated with a Gaussian model based on a constant
entrainment (or ‘‘turbulence'’) coefficient.
In the event of a release of a volatile liquid, vapourisation occurs rapidly. Generally, the rate of
evaporation is in the order of 10 m° hr'1m2 for a heat transfer coefficient of about 30 W m'2K'1,
where denser and more conductive floor material or lower normal boiling point increases boiling
rate (CCPS, 1996). Other influencing factors include air velocity and ground surface geometry.
Because of the rapid evaporation rates, dispersion of a vapourising pool is often treated as a
cloud release (Woodward, 1998). Kataoka et al (2000a, 2000b) supports this, where
concentration measurements following a liquid refrigerant release showed subsequent room
concentration to be almost homogenous.
1.2.5 Releases into rooms with ventilation

Several studies address releases into rooms given forced airflow which is analogous to the various
convection sources that cause turbulent motion within rooms; from external wind (i.e.,
infiltration), mechanical ventilation, and thermal convection from heat sources.
In addition to experiments under still conditions, Fletcher and Johnson (1986) also studied
releases o f CCty into a ventilated room whilst varying release location relative to air discharge
position. With small release rates and moderate ventilation rates, a homogenous gas concentration
was generally observed regardless of release position. Ventilation from floor upwards - C(in) to
X(out) in Figure 96 - and counter-flow to the plume flow when release at high level (Figure 96)
produced rapid mixing. Upwards ventilation with entering air at low7level and the outlet at mid
height - C(in) to Y(out) in Figure 96 - reduced the rate o f mixing. Downw ards ventilation
parallel to the plume direction - A(in) to Z(out) in Figure 96 - produced a homogenous mixture
from halfway between the release point and the air outlet. Similarly a release at mid-height gave a
homogenous mixture from halfway between the release point and the air outlet.
A higher mixing rate was therefore produced when airflow7is counter to plume flow but removal
of gas is more effective where a parallel flow configuration is used, although this leads to a higher
concentration within the layer. Fletcher and Johnson presented a modified decay-equation to
predict the change in room concentration, and included a factor for imperfect mixing between the
inlet air and the room air. This factor varied according to room size (for example it was shown
that mixing was more effective in smaller rooms), ventilation inlet location and airflow rate.
269
Recirculation fans were tested and resulted in near-perfect mixing. Also the effect of ventilation
when a stratified layer already existed within a room was investigated. Comparison with results
for airflow in parallel to the plume - e.g.. release at high level with A(in) and Z(out) in Figure 96
- shows the effectiveness of forced airflowr on mixing was diminished. This occurs because the
energy needed to overcome the inertia presented by the negative buoyancy of the dense layer is
greater.
airflow in release source airflow out
X(out)
mid Y(out)
low Zlout)
Figure 96: Positioning o f air inlet and outlet positions and release locations within a room
Effect of ventilation w as also assessed by Marshall (1984). In a similar way to Fletcher and
Johnson. Marshall looked at the effect of ventilation in relation to the release source location.
Where the two flow s were parallel - e.g., release at high level and A(in) to Z(out) in Figure 96 -
the concentration-height profile w as similar to the unventilated case, although more evenly
distributed within the layer. In all cases, higher airflow- rates reduced the maximum floor
concentration. Ventilation counter to the release - C(in) to X(out) in Figure 96 - resulted in a
similar concentration profile as was observed for the quiescent room, where two distinct layers
develop. Cross-room ventilation was also considered and measurement showed lowest
concentrations were obtained when the outlet vent was at low level (in the case o f a dense gas)
irrespective of the inlet position. Similarly with a dense gas, an outlet vent at high level resulted in
negligible mixing. The dependence of floor concentrations on release velocity7lessened with
increasing ventilation rates. In line with the findings of Leach and Bloomfield, it was found that
much greater ventilation rates w ere required to initiate mixing o f an already stratified layer, than
if the ventilation was operating during the release.
270
Leach and Bloomfield (1973) conducted similar mixing experiments with pre-mixed air and C 0 2,
i .e., with negligible buoyancy, and a second case with pure C 0 2, under a range o f ventilation
rates. Mixing of pure C 0 2 was less effective than with the neutrally buoyant C 0 2/air mixture
since the higher density' gave the release greater inertia. However, with a low release rate, a
moderately low air change guaranteed homogenous room concentrations. It was observed that the
position of the ventilation outlet affected the development of a stratified layer below the outlet
duct. A model was developed to describe the process, which showed good agreement with
experiments. Measurements also supported similar observations of a parabolic variation of
concentration with release height at low ventilation rates, but this concentration-height profile
approached linear as airflow' increased. The study also found that velocity of the discharge source
also affected floor concentration.
Heiselberg (1991) measured concentrations of C 0 2 for a constant gas flow' under a range of
ventilation rates. Increasing airflow gave near-homogenous concentrations. Also noted was a
peak in concentration at release source height. This result seemed inconsistent with other studies
but it can probably be explained by the C 0 2 and air being premixed prior to the release into the
room, which w'ould cany- it across the centre of the room, i.e., where the high concentrations were
observed. Above 1.5 h 1 air change, the concentration distribution w'as no longer influenced by the
increasing airflow rate. By varying the locations o f the air inlet and the source o f the release, it
w as found that these factors were important with respect to the density' of the gas. For example, a
ventilation source at low level when a high-densitv gas is released produced better mixing because
of the greater gravity induced convection within the room.
Clodic and Cai (1996) conducted tests with extract ventilation and air conditioning unit fan
operating. These showed significantly low er concentrations at floor level compared to the same
tests with no forced airflow. More recent tests by the same group (Jabbour and Clodic, 2002,
Jabbour et al, 2003) measured floor concentrations occurring from a release of different
flammable refrigerants with extract ventilation, unit airflow and vents within the room fabric.
These also demonstrated that presence o f higher airflow rates dispersed refrigerant more
effectively than without forced airflow. Positioning of mechanical ventilation ducts was changed
between the outlet at low and high level. Contrary to findings from other studies, these tests
show ed similar peak concentrations regardless of outlet position, especially when releases were
made close to the floor. Lastly, the effect of natural ventilation openings on the w all at floor level
resulted in lower peak concentrations, whereas elevated openings made no difference.
271
In terms of thermally induced convection, Cleaver et al (1994) reported that the air movement
generated by a domestic radiator was sufficient to prevent stratified layers occurring. As
described in CCPS (1996) and Woodward (1998), the effects of thermally induced convection
sources (e.g., solar radiation) increase mixing rate in the case o f outside releases. No studies were
found that specifically considered this aspect in confined spaces.

HCs (and other flammable refrigerants) are being employed in a variety of different RAC
applications and since this is relatively new the approaches used to address flammability hazards
are limited. Considering the areas covered by the literature review, it is evident that current safety
guidance neglects to account for many aspects associated with the behaviour o f flammable gases.
A number o f key areas relevant to the flammability risk were identified from the literature and
these should be accounted for when developing improved methods for addressing the safety
requirements o f relevant RAC equipment

There were a number o f subjects related to the flammability hazard of HC refrigerants and the
main observations follow.
- Risk assessment o f flammable gas and refrigerants. Most QRA methods are for very large
outdoor releases o f flammable gas, and have little relevance to flammable refrigerants.
However there have been a number of QRAs focussing specifically on the RAC situation,
and an overview of the way in which various parameters were dealt with is detailed in Table
35. For ignition probability, mostly a single value was used regardless of the environment,
and only differentiated between instant and delayed ignition.18 Between two and three
different leak sizes were assumed and data came from discussions with service people, or
using methods normally employed for industrial plant. Dispersion o f refrigerant was
accounted for by dispersion tests specific to the equipment being evaluated, or from basic
dispersion models mainly used for external releases o f large gas clouds. Similarly,
consequences including explosion and/or fire were mainly evaluated using methods developed
for industrial situations.
- Gas dispersion in confined spaces. Only a limited number o f published documents relate to
the specific conditions of interest, being finite releases o f dense gas into confined space with
various forms of airflow7. Those that are relevant generally demonstrate many similarities in
18 Instant ignition is where material is ignited whilst being release, and delayed ignition follow s
cessation of the release.
272
their analysis and conclusions, which provide confidence in terms of interpretation of mixing
phenomenon in confined spaces. Several papers also related specifically to refrigerants.
Mixing of a release is dependent upon many factors, and it is therefore important to precisely
define the conditions. This includes establishing the release as instantaneous or continuous,
ventilated or unventilated and aspect ratio. Generally in terms of a release o f dense gas, a
high velocity, elevated, upwards directed source would promote good mixing. Conversely, a
downward, low velocity release near to ground level will inhibit mixing, equating to higher
floor concentrations. Volatile liquid releases generate turbulence as they vapourise, which
normally mixes more effectively than a gaseous release. A dense release into a room mixes
with the ambient air, and upon reaching the floor develops a stratified layer. Disagreement
exists in characterising the concentration-height profile; some studies indicate two distinct
layers, whereas others demonstrate a continuous parabolic gradients, but probably due to
differences in release and environmental conditions. There is a clear relationship between
developed concentration, release rate and airflow rate. Above a certain ventilation rate,
homogenous concentrations are achieved. The direction of the airflow discharge relative to
the direction of the release is significant in terms o f mixing. The presence of additional
objects within a space will generally result in improved dispersion except where channelling
occurs to confine the release further.
It transpires that studies either focus specifically on earning out a QRA, or investigate gas
dispersion, and little attempt has been made to address both aspects simultaneously.

QRA evaluations are typically based on models for industrial and large-scale situations that lose
relevance when applied to non-industrial applications. The review highlights the necessity for a
more detailed approach to the evaluation of risks associated with the use of HC refrigerants, and
to establish more thorough guidance for design and installation practices and guidance relating to
the factors that affect gas dispersion.
The QRA approach is considered to be the most suitable for addressing the influence of design
and installation characteristics of equipment. The review indicated a more concise approach to
QRA in RAC equipment is required where a broader range of parameters need to be accounted
for. This includes a proper assessment of SOI, an improvement in evaluating leakage, and
thorough characterisation of dispersion of refrigerant. It is also important to account for different
operating situations, including cycling and safety controls. In addition to normal operation, the
273
method should also be capable o f handling servicing and storage activities. The work was carried
out in several stages.
Table 35: Summary o f the ways that main parameters were determined in ORAs
Reference Parameter
SOI Leakage Dispersion Consequences Stages
Goetzler et Probability7of Small and Experiments Based on Operating (for
al (1998) each being catastrophic, of floor ignition different
active anecdotal concentrations experiments locations
frequency for within a
catastrophic residence).
and small leak servicing
Wolfer and Constant Instant/ Decay Assuming Operating.
Seiler (delayed. constant. equation explosion or servicing
(1999) instant) catastrophic/ fire outcome
small according to
situation
Elbers. Constant 1%. 10%. Commercial Explosion Operating.
Verwoerd, (delayed. 100% of full ‘'Release” severity based servicing
van instant) bore rupture, model (mainly on commercial
Blanken. and decay model
van Gerwen frequencies equation) prediction
(1996) from
industrial
plant
ADL (1998) Constant Catastrophic Maximum Explosion or Operating.
(delayed, leak, small flammable fire severity defrosting.
instant) based leak. volume from based on servicing
on estimate anecdotal total release commercial
frequencies software
Probability models
Section 2 presents the development of the overall QRA model. The first step was to break down
the events that ultimately lead to ignition o f refrigerant and subsequent effects. This included
identifying failure modes and local environmental characteristics that affect the likelihood of
ignition and the severity of the resulting events. A general framework for calculating frequency of
ignition and its consequences was developed. Similarly, a set o f numerical limits were determined
in order to judge the significance o f the calculated risk, Individual probability models were
produced to calculate the components o f the overall risk; leak frequency, SOI probability and
probability’ o f a flammable concentration. Each o f these component probability7models relies on
physical data to represent the mentioned events.
274
Dispersion of refrigerant leaks
Suitable data on dispersion of refrigerant leaks was largely absent so it was necessary to conduct
suitable experiments, and use this to develop a dispersion model in Section 3. RAC equipment
and installations are reviewed to identify the characteristics that would influence gas dispersion,
and experiments were designed accordingly. Concentrations from simulated leaks were measured
and the data was subsequently employed to produce a number of correlations and incorporated
into the dispersion model.
Consequence
Ignition of a flammable mixture produces fire and explosions that determine the likelihood of
fatality and other consequences. Section 4 presents additional sub-models to calculate the severity
of the consequences of thermal intensity and overpressure to be employed by the fatality7
probability7models.
Evaluation of risk
In Section 5, outputs from the entire risk model are presented, and attention is paid to the
influence o f the various factorson calculated risk. An exhaustive examination o f these various
factors helps identify the effectiveness o f risk-reducing measures that can be applied to the design
and installation o f RAC equipment.
275
SECTION 2: GENERAL RISK ANALYSIS
2.1 Fundamental risk model

The purpose of this section is to analyse the sequence of events, and quantify the individual
events that lead to a hazardous outcome. Prior to any numerical assessment, a qualitative analysis
was carried out where the sequence o f events, from component failure to consequence, was
identified. The possible occurrence of each event leading up to the hazard must be interpreted in a
way that allows it to be evaluated numerically, thereby accounting for physical conditions that
affect the outcome of the event. Subsequently, probability and frequency equations must be
formulated for each o f the components leading to, and influencing the consequences. This also
requires the sourcing of empirical data suitable for applications within the scope of the QRA.
Start
Failure of pressure system | No failure of pressure system
leak frequency
Release inside Release outside
Gas doesn't enter control

Gas enters control volume
volume
probability o f
Gas dilutes to within Gas dilutes to outside ignition
flammable limits flammable limits (probability o f
SOI, probability'
of flammable
SOI in reference volume No SOI in reference volume volume)
SOI active SOI not active
ignition
Ignition No ignition
frequency
Fire/explosion No fire/explosion
consequence
Damage to people/property No damage to people/property
End
Figure 97: Sequence o f events leading toflammability hazard
276
The basic sequence o f events leading to an undesirable outcome is: component failure - creation
of a flammable mixture - ignition o f the mixture - damage. Figure 97 shows the sequence,
identifying the possible outcomes of each discrete event. Also indicated are the frequency and
probability calculation requirements corresponding to specific groups of events. The first output
of interest from a QRA is the frequency of ignition - expressed as the number of occurrence o f an
event per unit time, for a specific RAC equipment design, within a particular environment. With
reference to Figure 97, an ignition event is coincidence of three fundamental events:
- occurrence of a leak,
- development of a flammable-refrigerant/air mixture at a specific location, and
- co-existence of a source of ignition being “active” within the flammable mixture.
These events have their individual probabilities or frequencies, and the frequency of ignition is
essentially the product of these. The ignition frequency o f a single leak under any one set of
conditions ( ) is calculated from equation (266), where f leak ! l and Pf*t are leak frequency
and probability' o f ignition of a flammable mixture by an active SOI, respectively. is
evaluated with respect to a vertical element/region ( h ) within the room so that SOI are linked to
local formation o f a flammable mixture (in light of the relative buoyancy o f the gas), and f leakAl
is of a specific leak size in term o f duration ( / ) for a set of conditions.
3
(266 )
where / represents each set of operating and environmental conditions and h refers to low',
medium or high region within the room. f i eak l i is largely related to the construction of the sy stem
and compressor operating mode (on- or off-cycle), and is normally based on empirical (and
anecdotal) data. The size o f a flammable mixture (in terms of volume) and its duration is
influenced by many parameters, including charge size, leak duration and airflow conditions,
arising from infiltration, convection by thermal sources, evaporator and/or condenser fans, and
various forms o f mechanical ventilation. Thus PÂj is primarily dependent on environmental
conditions and equipment, and requires modelling of a release to determine relevant

characteristics, and physical data for the number, location and active time o f local SOI.
Description o f the sub-models used to quantify' f leakl, and P^*t are given later.
277
To account for the dynamic nature of equipment operation and/or associated environment, / J
was evaluated for each set of conditions. Figure 98 identifies the possible variations for each
parameter, and f*t was therefore evaluated for the combination that applied, such as compressor
operating mode and alternate or multiple airflow types. Consequently, the overall ignition
frequency ( / * ) is the sum of the individual ignition frequencies for all leak sizes, weighted with
the corresponding time fraction for each combination of conditions - compressor operating mode
( <f>op) and each airflow type - that is present (equation 267).
Sleak
/•= ! Z iZ /iW * , (267)

1=1 /=1 V i-1
with 4 for <f>op. <f>2 for for <f>thenn, (f>A for <f>umJ and (f>5 for <f>mech It is noted that for
RAC equipment that comprises a number of refrigerant circuits, the total ignition frequency for
the equipment is the summed / * for every circuit.
Normal
operation
Compressor Compressor
on-cyele off-cyele
Leak Leak Leak

size 1 size 2
Unit air Unit air

off
Mech Mech
vent on vent off
Therm Therm
sources off
Infil Infil Infil

rate 1 rate 2 rate n
High Mid Low

cloud cloud cloud
SOI SOI
present absent
Figure 98: Partial event tree illustrating variations o f conditions
278
Figure 98 shows that the compressor may be operating or not ( o p ), and for each operating mode,
there are a range o f different leak sizes. During any individual leak event, the dispersion of the
release is affected by air movement due to any combination o f unit airflow ( u n it), mechanical
ventilation ( m ech), thermal sources ( therm ) and some infiltration rate ( in fil). A subsequent
flammable mixture may be present at high, middle and/or low level within the room, where an
active SOI may be present.
Ignition of refrigerant may result in fire and/or explosion, where the subsequent heat and/or
pressure wave can cause damage and injury to occupants. For this study, intensity of thermal
radiation and overpressure (within the room) were used to evaluate consequence. Where the
consequence was of sufficient severity, the individual risk o f fatality (fatalities associated to the
installation) was estimated. Societal risk is not addressed specifically since for relatively small
releases the consequence normally affects only those within the room occupied by the equipment.
Frequency o f fatality ( f faSai) is the sum o f individual fatality frequencies resulting from each fj*
(equation 268).
(268)
i=i L /=]
where ^ and ^ CI are the number o f occupants and their residence time fraction,
respectively. Pfata} (/,* ) is the probability of fatality due to thermal intensity ( I ^ ) and/or room
overpressure ( h p °Rm), resulting from ignition under the conditions corresponding to /,* .
In order to represent the overall risk impact posed by an installation, the use of integrated thermal
intensity-frequency (equation 269) and integrated overpressure-ffequency (equation 270) were
proposed and used. They represent the time- and severity-weighted consequence o f ignition.
( Nleak
(269)
j=i V /=i
( Nleak
(270)
For convenience %IOCC and were combined to form an overall dimensionless "‘risk factor”
(equation 271). The formulation of equation (271) was chosen simply to account for the
combined damage or injury caused by both consequences occurring simultaneously.
279
X = >/(fl, Z i ) 2 +(<*2 - Xspo) 1 (271>
where constants a, and a-, are included to provide a weighting for each type o f consequence.
Determination of ax and a2 was considered to be largely subjective as different parties may rate
one form of injury less significant than another. Here, the values ax = 0.075 s'1 (kW 1 m")4 3 y and
a2 = 1 kPa'1 y w-ere chosen because 0.075 represents the relative difference between the
acceptable limits of XIocc and Xspo ^ discussed in the next section.
2.2 Context and risk acceptance criteria

It is necessary to put the calculated risks into context if equipment is to be judged as ''safe” or
‘‘unsafe’’. This was achieved by establishing a set of limits to compare against the outcome o f the
QRA. indicating an acceptable or unacceptable risk considering the following measures:
- frequency o f ignition.
- frequency o f fatality,
- severity- o f consequence, and
- integrated consequence-frequency, or “risk factor”.
Ignition risk was interpreted as fire risk, since ignition may lead to a secondary fire; as discussed
in Section 4, addressing the likelihood of a secondary fire from ignition is complex, so frequency
of fire is equated directly to ignition frequency. It is useful to compare the risk o f RAC equipment
against risks from other household and small commercial appliances. Based on statistics of fires
in the UK (Collier and Watson, 1998) and product population data (Waterston, 2003), the
following frequencies of building fires originating from an appliance, were obtained: refrigerators
(non-HC) - 1.13><1 O'5 v 1, gas cookers - 8.71 * 1CT4 v*1, gas central heating - 4.03><10'5 y'1,
washing machines - 1.31 xlO-4 y’1. televisions - 2.65 xlO'5 y'1, and other audio/visual - 8.08x 1O'6
y'1. Causes were electrical faults or a release o f flammable gas depending on the type of
equipment. These fire risks are evidently considered “acceptable” by society, and therefore may
be used as a reference point for acceptability of the ignition risk o f flammable refrigerants. The
lowest (rounded) value for all the appliances listed in the data w as used as the acceptance limit
(Table 36), so it is favourable for f * to be lower than these appliance fire frequencies.
In terms of severity- of a consequence, limits corresponding to those that could cause a fatality'
w ere chosen. Alternatively, a lower value based on some degree of injury or physical damage
could be adopted. For fatality' risk, the Health and Safety- Executive (HSE) recommend values for
280
“negligible" risk to an individual: member o f the public f fatal = 1x 10-6 y'1, and service industry7
worker f fatal = 1* 10'5 y'1(HSE. 2001). Calculated f fatal should be lower than that specified by
the HSE (Table 36).
Similarly, the acceptance limits for Xiocc ■

> X\po ^ d X were derived according to likelihood of
fatality and the tolerances specified by the HSE.x locc and x^poare conceptual parameters, and
it is not possible to directly adopt empirical values for the acceptance limits. Instead, limiting
values of Xiocc ^ d X&po were determined from the lowest integrated consequence-frequency, as
derived from the limits for fatality frequencies and the probability of fatality associated with a
chosen consequence. From equation (269) and (270), the formula for Xiocc ^ d Xapo can
reduced to equation (272) and (273) by assuming one single leak size and a single set of
conditions.
Xjocc= r - I occ (272)
X a Po - / * ‘ N?°Rm (273)
The frequency o f fatality for a given consequence is the product o f ignition frequency and the
probability7of fatality associated with a given consequence (equation 274).
f Ma, = r -PfMal (274)
Equation (274) can be transposed for / * , and then substituted into equation (272) and (273), to
give equation (275) and (276).
X ^ '^ - l o c c (275)
fatal
(276)
fatal
Pfatai can be calculated from the severity of a chosen consequence as detailed later in Section 4,
which is normally7achieved using a logarithmic formula. Replacing Pfatal in equations (275) and
(276) with the appropriate formula gives equations (277) and (278).
281
where the constants a and b are specific to each consequence.
Using equations (277) and (278), acceptance limits for x locc and %Apo can be determined
according to a chosen value of f fatal, which as previously mentioned was taken from HSE
(2001). The severity of the consequence - I ^ and hp°Rm - necessary to cause fatality varies,
and since equations (277) and (278) imply non-linear dependence of Xiocc ^ d Xspo on the
consequence it was necessary to evaluate over an appropriate range of I ^ and &p°Rm. For
thermal intensity, the lowest value that can cause fatality is I ()CC ~ 1050 s (kW m 2)43 and the
lowest overpressure is &p°Rm - 2 5 0 kPa, and so equations (277) and (278) were evaluated for a
series of incrementally larger I ^ and k p Rm. Subsequently, the smallest value of Xiocc ar|d Xspo
calculated from the range of 1^ and ApRm was selected as the acceptance limit.
Table 36: Maximum acceptable risk criteria
Criteria To members of public To sendee industry workers
Frequency of ignition (fire) ( / * ) < ixkt V 1 < lx io - V
Thermal intensity ( I ^ ) < 1050 s (kW m 2)4'3 < 1050 s (kW m'2)4/3
Overpressure ( &p°Rm) < 250 kPa < 250 kPa
Frequency of fatality ( f fatal) < lx lO ^ y 1 < lxlO'5 y'1
Integrated thermal intensity-freq.

< 5.6xlO*3 s (kW m'2)43 y 1 <5.6xl0‘2s(kW m '2)4/3y 1
( Xloee )
Integrated overpressure-freq. ( x^po) < 4.2 xlO*4 k P ay 1 < 4.2xlO'3 kPay"1
Risk factor ( X ) <5.9xl0'4 <5.9xl0‘3
Using f fatal for ‘‘members of the public”, the minimum Xiocc was found to be 5.6 x 1O'3 s (kW m'
2)43 y 1, which resulted from 1occ ~ 3000 s (kW m'2)43, and the minimum X apo was 4.2x 10“*kPa
y'1 corresponded to Ap^m ~ 330 kPa. The limits are included in Table 36. and equivalent values
for '‘sendee industry worker” is also listed. Lastly, for calculation of the overall risk factor X ,
the value of ax in equation (271) is obtained from the ratio %hpo ! Xiocc = 0-075. Evaluating
equation (271) when ax = 0.075 s’1 (kW 1 m2)4^ y and a2 = 1 k P a 1 y, the acceptance limit of X
for both activities was calculated and listed in Table 36.
2.3 Leakage frequency

Reviewed QRAs employ leak frequencies which v an by size o f leak, activity and equipment type.
A compilation of leak frequencies from the literature (Table 56, Appendix 1) show values
extending from 3* 10'9 y'1 to 2x10'’ y"1 for normal operation. Some studies adopt a single
frequency to represent all leaks, whereas others differentiate between small and large leak sizes.
Reiving on these reported values for the QRA model was not deemed appropriate due to the
inconsistency between equipment types, but more so, because most were simply based on
interviews with service technicians and manufacturers. Further, considering the strong
dependency of risk upon leak characteristics (as detailed later), it is important to account for a
variety* o f releases anticipated for a given equipment construction. This requirement led to the
development o f a new leak frequency model.
The model employed leak data ranging from the smallest leak mass flow rates to largest
(''catastrophic”) leak, and the data suited to the equipment under consideration in terms of
number and types of connections. Statistical data on leakage is currently scarce; however,
permanent refrigerant monitoring is employed on systems for leak detection. This type o f data
was considered a useful source for characterising leak frequency. A paper by Radford (1998)
provides detailed information on such measurements, upon which the frequency model was
derived.
There are two possible events for a system that contains refrigerant; the system leaks ( E leak), or
it doesn't. The probability of E leak is the ratio o f leak events, to the total events within the set
( S ), which comprises leak events and non-leak events (equation 279).
= (279)
N(S)
The number o f events ( N ) can interpreted either on a time basis or a leak initiation basis and it is
important to differentiate between these two definitions. A time-based probability can be related
283
directly to mass flow rate of a leak and therefore this approach was employed for dispersion
calculations (whereas an initiation-based leak frequency cannot). Equation (279) is presented in
terms of the ratio of time that a system leaks to total time (equation 280).
(280)
tot
If the total time ( ttot) is set at one year and the duration o f the leak ( tleak) is within the same
period, the leak probability7is a frequency ( f leak). When equation (280) is applied for a range of
leak sizes (assumed to correspond to a selected range of measured concentrations at specific
locations. C ), f leak of a given leak size per connection (or joint) can be obtained from equations
(281) and (282).
(281)
conn
(282)
rconn r'
where N conn is the total number of the connections, N E leak (C ) is the number of leak events that
result in a measured concentration range per year, and AtE is the event interval time. A further
consideration is the shortest time of a catastrophic leak w here the rate of refrigerant release is
largely independent of leak aperture; the flow rate is dictated by internal characteristics of the
system. Thus, equation (281) does not necessarily apply in some cases and instead tleakcat is
determined and fixed based on experiment. Further, f leak differs according to the compressor
operating mode, so separate values for on-cycle and off-cycle conditions are required.
As discussed earlier, a large number o f leak sizes increments ( N leak) need to be evaluated from
smallest to catastrophic leaks. In order to obtain a smooth variation of leak data over the range of
chosen leak sizes, the empirical f kak and tleak was correlated in the form of polynomial equation
(283).
2
J leak,I = C t\ * ^ leak,I + ^ leak,I + (283)
In determining ax, a2 and a3 in equation (283) direct polynomial fitting cannot be used due to
the fact that the number of incremental leak sizes assigned ( N leak) may be different to the number
o f leak sizes in the empirical data. In order to evaluate the frequency at any other number of leak
284
sizes, the coefficients need to be determined specifically for the selected number of leak size
increments. The procedure adopted is detailed in Figure 99. Empirical data for leak frequencies/
times (presently, for small, medium and catastrophic sizes) was adjusted by using an assumed
coefficient b (in equation 284), giving: f leak l = b • [ f UakJ ]meas and tleakl = b ■[fuakAmea*' *' >
except with a catastrophic leak, when t kakJ - tleak cat. Once the data is fitted {ax, a2 and a3
found), it must be checked to ensure the total leak time o f all leak sizes is equal to the empirical
leak time, using equation (284). Equation (284) is evaluated for the range of tleak, and if not
satisfied, b is adjusted, and the process reiterated.
M ea t M e a t- 1
y , ^ leak,} ^ leak,cat ^ ^ ^ \ f leak,1 (284)

/=1 /=1
Note that since Ph ,* is a function of rteai. ; therefore comparison o f f ki for different operating
modes would be valid onlv if identical incremental leak rates are used.
Choose inputs: , initial guess for b(Nleal. ) = 3 / Nleak
_______________________________ i ______________________________
Calculate small, medium and catastrophic leak frequencies using empirical leak
frequencies and N co^ e.g., for small leak size: (srnl) = (sml,Nconn = 1) • N com
Similarly, calculate small, medium and catastrophic semi-empirical leak times, e.g., for
small leak size: tlak(sml) = fZat(sm l)t'
Calculate, total semi-empirical: (tot) = _/£* (sml) + f ^ (med) + f j ^ (cat)
±________________________________
Adjust semi-empirical leak frequencies and times according to the selected number of
leak sizes ( ) , e.g. for small leak size: (sml) = b(Nleak
'L* C5™/) = fkak(•””/) • ’ although catastrophic leak time is fixed, i.e., (cat) =
_______________________________ f __________________________
Determine constants, ai, a2, and a3 for quadratic equation using the adjusted semi-
empirical and t ^ for small, medium and catastrophic leaks.
Choose incremental leak times for QRA, , and use quadratic equation to calculate
corresponding incremental leak frequencies, f le a k , ,
and use to calculate the sum of
Nleak ’ j
incremental leak frequencies: (tot) = y . fle a k J
/=1 i
Yes
Adjust number leak of sizes coefficient:
Figure 99: Calculation sequence to determine adjustment fo r selected number o f leak sizes
285
The leak frequencies used in the model are subject to a final adjustment to account for the locality
of leaks. To address leaks that only occur from a portion of the system within a specific location
- say indoors - the values obtained from equation (283) are then tailored as equation (285).
f leak ~ f leak ' Plocal (285)
where Plocal is the probability o f the leak occurring locally. Considering an occupied room, Ploca]
would be unity for integral equipment and <1 for split-type equipment.
For application of the model (equations 281 to 284) it is necessary to obtain empirical values.
Data on refrigerant leakage was made available (Ayers, 2000), which had been gathered from
monitoring 410 supermarket installations throughout the UK. Refrigerant concentrations were
measured at 30-minute intervals, and quantities leaked and source locations recorded. Out of the
108 million samples, 1,533,239 leak events were detected and measured; 1,435,473 were
regarded as 'small' leaks, 86,882 were 'medium' and 10,884 were ‘catastrophic' (categorised
according to measured concentrations). Using equation (282), and taking the average number of
connections per supermarket system as 700 (excluding those on pre-fabricated components such
as coils), correspondingly leak frequencies per connection are: small leaks, f leak (N conn= \) =
2.85x 10“*y 1, medium leaks, f i eat(N com =1) = 1.73* 10"5 y 1; and catastrophic leaks,
fieak (Ncom, - 1) = 2 .16x 10-6 y*1. To apply these derived frequencies to a system o f specific
design, they are multiplied by the number o f connections ( f Uak = fleak (N conn = 1) • N conn).
It is believed that adoption of these data addressed some of the concerns highlighted for existing
frequencies: it is based on measurements, and given the large sample set it accounts for a wide
range o f installation- and service- quality7and a large number o f different types of system
components and connections. It is noted that further work would enable adjustment of f leak
according to the vulnerability of different components to leak.19
These data, particularly for catastrophic leaks, apply only to systems that have their compressor
operating (on-cycle), since none o f these installations completely shut down. A system in off-
cycle has a catastrophic f kak less than on-cycle because many mechanisms that cause leakage
are not present, such as vibration, pressure and temperature changes. The only situation found to
19 The approach in Section 2.3.1 of Part 1 could be extended for weighting leak frequencies according to
the likelihood of leakage from different components, for example, flared versus brazed joints.
286
have the potential for failure during off-cycle was a combination of fatigue and stress corrosion
cracking under severe environmental conditions (Guvoncourt and Fennell, 2000), leading to an
estimated catastrophic leak frequency of 1x 1O'8 y'1 for a split air-conditioner, or f leak = 1)
~ 1x 10'10 y 1. The former is comparable to the historical figure o f / feajt = 3.Ox 10'8 y'1
recommended for LPG tanks (Roberts, 1980), which was considered analogous. The above
supermarket measured frequencies for small and medium leaks were assumed the same for both
on and off-cycles since leak duration is typically longer than typical cycling times. In general, off-
cycle f leak appears to be about 1(F lower than for on-cycle. Referring to equation (285), analysis
o f the sources o f leaks from Radford (1998) and Ayers (2000) show that 28% of the leaks
occurred from components normally located indoor (e.g., expansion device, etc), so PkaL. =0.28
to account for indoor leaks from "split" equipment, otherwise for integral equipment, Pleak = 1.
Alternatively, Goetzler et al (1998) assumes a 50% split between indoor and outdoor parts, i.e.,
p1 leak = yo j 5j -20
Servicing f leak is greater than for normal operation since the technician could be intentionally
breaking in to the system. Goetzler et al (1998) recommended catastrophic f leak - 1 Ox 10-3 y'1
for equipment normally having 20 to 30 connections (Table 56), thus f leak (N cona = 1) = 4.0x 10-5
y'1. This frequency is about 20 times greater than that o f normal on-cycle operation.
Alternatively, comparing the data in Table 56 shows that servicing f kak is typically between 1
and 3 times that o f total f kak for normal operation. Thus, small and medium f kak were assumed
proportionally higher compared to normal operation.
2.4 Probability of ignition

In the review of QRAs, probability o f an active SOI surrounded by a mixture at a flammable
concentration ( C F ) normally involved two fixed values: one for immediate ignition and one for
delayed ignition, e.g., 8x 10-3 and 2x 10-3 respectively (Blom-Bruggeman et al, 1996a). This is
typical o f industrial risk assessment, where single generic values are used because the exact SOI
characteristics for large areas of land are unknown. The probability of immediate ignition is
applied to ignition during the release, which typically results in a less severe consequence than
delayed ignition. The exception is Goetzler et al (1998) who employ a constant probability
specific to each location o f system parts, betw een 1x KT4 and 2x 10'2. Woodward (1998) detailed
20 Again, the approach in Section 2.3.1 of Part 1 could be employed to determine this probability more
precisely.
287
the model used in a HSE code, which is for a flammable cloud passing over urban, rural or
industrial land (equation 286).
P F* = l - ( \ - P F)A!AF (286)
where P F is the probability o f having a local C F , A is the total area covered by the cloud at
some point in time, and A F is the total area covered by the portion of the cloud at C F , until it
disperses to below LFL. This approach is not applicable for releases into rooms; considering the
case of a flammable cloud that entirely covers the room floor, P F * = P F = 1. For a cloud of
varying volume and duration this is counter-intuitive, and further, it cannot account for different
populations and characteristics of SOI. A more suitable approach was offered by Rew et al
(2000), being location and time specific. P F* (at a given time) is predicted on the assumption
that the likelihood of active SOI increases over time consistent with expansion of the cloud
(equation 287).
f
Pr>F
. t*SO1
SOI
\ ( pF
soi "
*
rO
^
P F*( 0 = 1 - 1 •exp (287)

At + t SOI J
where Psoi is the probability o f the presence of an active SOI within a potentially flammable
cloud, t r0 is the time since initiation of the release, t soj is the active time of the SOI, and At is
the average time interval between active SOI events. Rew et al give empirically derived values for
PF
m, t soi and At according to different locations and time of day. Whilst this approach better
account for the local situation compared to equation (286) and values are provided for selected
indoor locations, it still lacks the ability- to account for SOI specific to a room, and it is vague in
how7it handles the duration of the flammable volume.
To address the shortcomings of the methods currently in use (with respect to the present purpose),
a new formula was developed where the characteristics of each individual SOI and each release
scenario may be accounted for. For a given space-time, ignition could potentially occur when an
active SOI is present at the same time and spatial position as C F . To illustrate the quantification
of P F*, the Venn diagram in Figure 100 shows the existence of a potential SOI event, the active
SOI event and the C F event. The coincidence of these leads to ignition, as indicated by the
shaded area. To determineP F*, it is evident that two component probabilities are needed:
probability7of an active SOI at the same time as C F , and probability7of C F at the same spatial
position as the potential SOI.
288
ignition
active
SOI time
volume-time of room
Figure 100: Venn diagram indicating occurrences necessary for ignition
Occurrence of SOI events

Within most domestic and commercial occupied spaces, an active SOI would be an electrical
spark of at least 20 mJ and a flame or surface at a temperature of at least 460°C. SOI may be
present on the RAC equipment or elsewhere w ithin an occupied room, a storage room, and tools
used for servicing activities. For a single active SOI ( Esol) assumed to occur for some duration
randomly within a reference period ( S ), the probability can be given by equation (288).21
(288)
N(S)
Similarly, if C F occurs randomly over a given duration ( E F ) within the same reference period
(.S'), the probability of C F being present is equation (289).
P{EF) = N(<E - (289)

N(S)
As E soi and E F are independent, the probability occurring simultaneously is the sum of both
probabilities (equation 290).
P(EL)=m
*" N(S)
d +^
N(S)
l <290)
21 If knowledge of the temporal behaviour of the SOI and/or flammable concentrations is known,
alternativ e methods can be applied to obtain a more precise representation of probability.
289
For each potential SOI, the number of events N (E sol) is represented by the active time of a
single event within the reference period, and N ( E F ) is the time in which C F is present. It is
assumed that this flammable time ( t F ) is continuous, i.e., the leak produces constant flow until
the refrigerant charge is depleted. Therefore, equation (290) can be wTitten as equation (291),
showing the probability of an active SOI within a flammable volume within a given reference
volume under a given set of conditions resulting from a certain leak size.
r t soi + tF
l.h.i A
Psoi,h,l,i - P avail (291)
t'
\
where P fokKU < 1, and t' is the reference period. P ^ , is included to account for the availability
of the SOI. For example, electrical components on equipment may be protected against the
ingress of refrigerant ( P ^ , = 0), although for various reasons the integrity of the protection may
fail (0 < Pavail < 1). Otherwise, for permanent SOI, = 1. In general, active SOI events
arising from faulty electrics are neglected, and it is assumed that any active SOI is immediately
available to C F , i.e., no time lag occurs as gas migrates through housings, etc. Also, an external
fire is not considered as a SOI because ignition of refrigerant w ould make a negligible
contribution to the existing consequence (Ritter and Colboume, 1998).
Availability of any SOI is largely dependent upon its purpose. Electrical components associated
with RAC equipment should be protected according to safety standards, but a recent survey of
electrical components in hazardous areas of petrochemical plants (Murdoch, 2003) revealed that
14% of protective enclosures were ineffective due to degradation and poor repair, thus Pmail ~
0.14. In addition, for servicing activities, Goetzler et al (1998) state that 15% of visits require
brazing activities ( Pavail =0.15) and 40% o f service technicians may smoke ( PmM = 0.4).
Further, servicing activities may sometime be carried out by technicians who may not be suitably
trained in handling flammable refrigerants. This implies a difference in use of tools that have
SOI, therefore should be altered accordingly.
It is also noted that leakage data from Ayers (2000) showed that once initiated, leaks rarely
produce constant release rates, particularly with small leaks. Variations in operating and
environmental conditions - typically cyclic temperature and pressure changes - terminate and
resume leaks owing to contraction and expansion of joints, etc. As such, t F is likely to be
intermittent and therefore P F

oi would realistically be smaller than equation (291).
290
O ccurrence of flammable volume events
The probability of having a flammable volume ( P F ) is based on the coincidence of the active
SOI being present in the same spatial location as C F . As with P F

oi, most of the literature does
not define Py specifically, but use a probability based on the likelihood of C F occurring
following a release. Again, this approach is not sufficiently precise since any release will always
produce C F at some location within the space: it would only be relevant if linked directly to a
SOI at a given point in space and time. Instead, the approach adopted assumed the SOI are
located randomly within the same reference volume that C F may exist. To obtain P F it is
necessary to determine the total volume of the flammable mixture ( V F ) following a release (as
quantified by the dispersion model in Section 3 .3).
The probability model assumes two possible outcomes: a C F occurs within a given location
( E F ) representing a flammable volume, or it is free from a C F , whether it comprises a mixture
outside the LFL - UFL or is absent of refrigerant entirely. The probability of E F occurring is
the ratio of the number of flammable volume events to the total possible events for the flammable
volume to occupy ( S ), where the number o f events is interpreted as total volumetric elements
(equation 292).
P(Ef ) = ^ 1 (292)
1 N( S)
If the reference volume is discretised the number of events N ( E F ) is the number of elements at
C F , and N( S) is the total elements within the reference volume. The probability of the
flammable volume is therefore (equation 293).
wfiere the reference volume V corresponds to the horizontal levels chosen for grouping local
SOI ( V'h = ARm • hRm / N h), and V^} j is the flammable volume existing within that reference
volume. For the present study, three reference volumes ( N h = 3 ) were chosen because it is
convenient to identify SOI normally located at low7, middle of high levels. Following the initiation
of a release, VRl j changes over time, which w7ould demand an integral approach to determine
291
Py . Outputs from the dispersion model (Section 3 .3) show V F to increase somewhat linearly
until the end of the release, and is thereon approximately constant within the decaying cloud and
eventually falls to zero in a relatively short duration. This trend in behaviour o f V F was found in
most cases, and therefore it was considered appropriate to maintain the basic P F formulation
(equation 293) by employing an averaged flammable volume with respect to flammable time
( V F ).
Tw'O additional parameters must be accounted for, the proportion of systems that leak their
charge and the likelihood that an active SOI successfully ignite C F . Various sources report
system leakage data in the form of “annual leak rate”, which is defined as the ratio of total mass
leaked from a given population of systems to the total refrigerant mass within that population, or
“refrigerant bank*’. This ratio is interpreted as the probability of a system to release its charge
(P^ ). Failure to ignite a C F is referred to as “percolation" (Rew7and Spencer, 1998), and is
explained by the presence of small pockets o f unmixed gas or air within a cloud. Including P^v
and percolation probability ( Pperc) the probability o f flammable volume becomes (equation 294).
( f7F
pF - * h.l,i
- P s y s -P p e r c (2 9 4 )
ARm hh
P ^ is normally determined for each equipment category7, e.g., small integral commercial
refrigeration units, or air-conditioning chillers.22 P was estimated from Jabbour et al (2002)
where 48 tests were conducted to try7to ignite R290 and R600a whilst at C F in a small room.
Despite measurements showing C F at the SOI position, 27% of the attempts did not result in
ignition, thus P ^ =0.73.
Whilst V F is partly dependent on refrigerant charge, the actual final leaked amount may be less
than the total charge because a proportion is retained within system. Therefore, the mass used to
calculate various flammable quantities is the system charge reduced by the appropriate
proportion (y/). Measurements of catastrophic R290 leaks (PR 100374, 2000) show?ed the mass
released was less than the charged amount, and was due to vapour residing in the system and
" Evaluation of this is addressed in detail within Section 2.3 .1 of Part 1, which can be employed to
determine P according to system design.
292
being absorbed in compressor oil. After a 3 .5 minute catastrophic leak. 30% of the charge was
retained under static conditions or off-cycle, and on average, 18% for on-cycle; so y/ - 0.30 and
y/ =0.18 respectively. Additionally. Kataoka (1999) made similar tests by measuring the mass
of R22 removed from an air conditioner using a large recovery machine. The system contained
1.0 kg of R22 and 0.4 kg of mineral oil. A catastrophic leak was simulated after cessation of the
compressor, the system was found to retain 5% of refrigerant. Conducting the same test but with
leaving the refrigerant to soak for increasing longer time before the leak was simulated, show7ed
the retained mass proportion increased linearly, to 15% after the system had been static for 20
hours. The differences in values between PR 100374 (2000) and Kataoka (1999) may be
attributed to system internal volumes, and different solubility o f R290 and R22 in mineral oil.
Occurrence of SOI and flammable concentration
Lastly, it is necessary to combine the SOI and V F probabilities to determine the ignition
probability indicated in Figure 100. To simply the derivation, it is convenient to employ the
compliment of the preceding events: probability of no flammable volume ( P F = 1 - Py ), and the
probability of the SOI not being active within the flammable volume ( P Ft - P F • (1 - Pôi) )• The
probability o f no flammable volume and the SOI not being active within the flammable volume is
the sum of P F and P F

oi (equation 295).
P y \u ) + PvAU • (1 - Psoi,h,l,i ) (295;
where Py hJ i and P î h l i are from equations (294) and (291), respectively. Conversely, the
probability of having a SOI being active only within the flammable volume is the complement o f
Phj j (equation 296).
(296)
It is expected that most SOI would produce several active events within the reference period, t ' .
It is held that any P F*f event is independent o f the others since any active SOI event is random
and therefore occurs irrespective of any other SOI event. Hence, the total probability7of all active
events for a single SOI is the complement of the product of all non-ignition events associated with
the given SOI (equation 297).
(297)
293
Substituting equation (295) into (297) gives equation (298) for the total number of active events
( N e ) for each individual SOI.
(298)
Within any control volume, there may be several different SOI, o f which some may be identical to
each other. The total probability of all SOI is the sum of the individual probabilities for each
active SOI (equation 299).
Nsoi
<29
A'=l
This probability' is then used in equation (266).
2.5 Time fractions

Time fractions represent the proportion of time that the occupancy, operating mode and airflow
types occur as the ratio of residence time to some reference time (equation 300).
(300)
where tres is the annual operating or residence time for the parameter under consideration at
condition i . For compressor operation ( (f>op ), occupancy ( ) , thermal sources ( <f>therrJl), forced
airflow from the RAC unit ( <fium!) and mechanical ventilation ( <f>mech), tres is based on the
operating or residence times, top, tlherm , tuml and tmech, respectively.
Unit airflow or mechanical ventilation may be initiated by a controller that responds to a

refrigerant detector or pressure transducer. Whilst these control methods are operating as
intended, the residence time is effectively the reference time, thus tres = t ' . As far as V F and l F
are concerned, these airflow are always present, whether they are operating constantly or started
part-wav through a release and thus for a given set o f conditions, <f>unit i or t f ) ^ ^ = 1. However,
such unit airflow' and mechanical ventilation are subject to unexpected failure due to mechanical
faults or incapability' to detect refrigerant, and this was accounted for by the failure probability
( Pfail) and subsequent down-time ( t fail), i.e., the time taken for completion of the repair and the
airflow' resumed. Pfail is based on the number of failures o f components leading to cessation of
airflow', out of a total population within the reference time. Thus within t ', the average time that
294
the airflow is absent is P fajl x t fail, and the total time that the airflow is present is therefore
equation (301).
P e s .i fe s (())j P fa il A ^fa il, i (301)
where tresi0) is the intended operating or residence time for the airflow. In general, the cause of
airflow failure may include duct blockage, fan motor burn-out, faulty positioning or degradation
o f refrigerant detectors, or malfunction of electrical circuits or controls. Evidently Pfail can vary
widely depending upon the type o f equipment, manufacturer, quality of installation, etc. Based on
a limited number o f sources, example data is provided in Table 37. tfaiJ depends upon the degree
of inconvenience (if at all) to those associated with the equipment. For example, failure o f airflow
from an office air conditioner during hot weather may result in a short downtime (of days),
whereas a faulty' refrigerant sensor may only be corrected at the next inspection (possibly months
later).
Table 37: Information on mechanical component failures
Component Description Probability Source

13% per annum of gas appliance
Blockage of flues, Davies and Pearson
o-j
ii
flues - probability of a blocked RAC

o
o
ducts, etc. (1999)

unit duct assumed about half.
Fan and fan motor Air conditioning m anufacturer
P f* =0002 Aloisi (2000)
failure reported 0.2% per year.
Failure of some part
Air conditioning m anufacturer
of electrical/control P fo „ =0.050 Aloisi (2000)
circuits. reported 5% per year.
Failure of detection Statistics on failure of smoke

equipment to detectors (except where batteries P m =0.138 Collier and Watson
initiate airflow, had been removed) show 13.8% (1998)
alarm, etc. failed to detect for various reasons.
Another implication of unit airflow is its functional link to the compressor operation. For
example, if the evaporator or condenser airflow fails, low or high pressures can be generated,
which may cause the compressor to cut out. To account for this, calculation of (f>op employed
equation (301) to calculate tres, where tres((j) is the intended compressor operating time, and P fa il
and tfai! are those for unit airflow.
295
Infiltration rates vary over time, so evaluation of requires attention. For a room, infiltration
rates ( A infil) were calculated from the fabric leakage ( ) at the standard reference 50 Pa air
pressure differential ( A p ). As local weather conditions change over the year, different Ap across
the room fabric occur, leading to a range o f A infiI and an associated range of (f>mfil. To calculate
(f>infil, the corresponding t res must first be found and this is based on the frequency o f weather
conditions leading to each A p . The duration of the different weather conditions resulting in a
specific Ap is the sum of the wind speed and outside temperature time increments that lead to
that A p . An A mfil is calculated for each A p , and the range of these (between minimum and
maximum A mfil) is divided equally into a number of sub-ranges ( A A infil). Thus <f>infil for each
incremental AA infil, is based on the time that weather conditions that lead to each AA infil are
present, tres( A mfil) , (equation 302).23
^ ^res j infil,i )
Yxnfili ~ (302)
For a given room, A mfil was calculated using the method detailed in BS 5925: 1991. First, the
total pressure difference between the internal and external wall is required ( Aplol), being the sum
o f the stack pressure difference ( Apstack) produced by the differential air density between inside
and outside, and the pressure imposed by external wind ( Ap Mind) (equation 303).
= & P Stack.! + A P w in d j (303)
Apstack was calculated using equation (304), and Apwmd with equation (305).
(304)
(305)
where hRm is room height, u Mnd is wind speed, p aout and p a m are air densities and the wind
pressure coefficient A w a s fixed at 0.75 for convenience.
23 This approach is similar to the “bin” method used for ambient temperatures.
296
Table 38: Frequency o f mean monthly temperatures in West London
Month Jan Feb Mar Apr May June July Aug Sept Oct Nov Dec
Temp. (°C) 3.7 4.3 6.5 9.4 12.5 15.8 17.4 17.0 14.9 11.5 7.2 4.8
Freq. ( y 1) 0.085 0.077 0.085 0.082 0.085 0.082 0.085 0.085 0.082 0.085 0.082 0.085
Both outside air density (due to temperature) and wind speed vary independently throughout the
year. Knowing this, the year is discretised and Apto! is calculated for each range o f w ind speeds
for each monthly temperature. Mean monthly temperatures for one UK location, and
corresponding frequencies are provided in Table 38. Data in BS 5925 is given for the cumulative
proportion o f time that the wind speed differs from the average. A polynomial w as used to fit the
data for inland areas of the UK, which w as then differentiated, so that equation (306) could be
employ ed to quickly estimate the yvind speed for a given range o f time fractions.
<f>(u')= a x + a 2 ■u' + a3 ■u ,d (306)
where <f> is the time fraction, and =-0.3361, a 2 =-0.5308, and a3 =0.3156. u is the yvind
speed relative to the time-averaged ( (f> = 0.5) w ind speed (equation 307).
u , _ U wind(0) (3 0 7 j
11Hind
yvhere u Mind is the mean {(f) = 0.5) annual yvind speed (1.8 m s'1 for West London).
0.3 - r —
^ 0.2
0.1
0.0
X V
O is
0© (N fS
Infiltration rate, (h-1)
Figure 101: Time fraction fo r different in filtration rate ranges
297
S e le c t in g a f a b r ic le a k a g e (in t e r m s o f ), e q u a tio n ( 3 8 2 ) w a s u s e d t o e s t im a te A mf]l fo r th e
r a n g e o f c a lc u la te d Apm . B a s e d o n a r a n g e o f c a lc u la te d A ]ntil fr o m m in im u m t o m a x im u m
v a lu e s , a s e t o f e q u a lly s iz e d AAjnfil e le m e n t s w e r e d e te r m in e d . T h e tim e fr a c t io n o f e a c h AAinfil
w a s fo u n d b y c a lc u la tin g th e to ta l tim e ( tresA(A in fiU ) ) th a t a g iv e n o u ts id e te m p e r a tu r e a n d w in d
s p e e d w e r e p r e s e n t, th a t le a d s t o e a c h AAmfil (e q u a tio n 3 0 2 ) . A s a n e x a m p le , a A ^ o f 1 4 h '1
w a s s e le c t e d a n d f o r t e n in c r e m e n ts , AAinfil w a s c a lc u la te d fo r a 4 0 n r * 3 m h ig h r o o m . T h e
r e s u lt s a r e p r e s e n t e d in F ig u r e 1 0 1 , s h o w in g t h e lo w e r AA infil ( 0 . 4 h ' 1 t o 1 . 6 h 1) o c c u r m o s t
o ften .
298
SECTION 3: DISPERSION OF REFRIGERANT LEAKS
3.1 Dispersion experiments

T h e p r o b a b ilit y c a lc u la t io n s in S e c t io n 2 . 4 r e q u ir e t h e d is p e r s io n o f a r e le a s e t o b e b o t h s p a t ia ll y
a n d te m p o r a lly c h a r a c te r is e d . F o r th e r a n g e o f c o n d it io n s r e le v a n t t o th is s tu d y , th e d a ta a v a ila b le
fr o m e x is t in g lite r a tu r e w a s in s u f f ic ie n t in s c o p e a n d o f m in im a l p r a c t ic a l u s e . T o r e m e d y t h is , a
n u m b e r o f e x p e r im e n ts w e r e c o n d u c t e d t o o b ta in d a ta s u ita b le fo r th e d e v e lo p m e n t o f a d is p e r s io n
m o d e l b a s e d o n a p u r p o s e b u ilt t e s t c h a m b e r , w h ic h in c lu d e d s o m e in itia l s m o k e v is u a lis a t io n
t e s t s . T h is s e c tio n d e ta ils e x p e r im e n ta l w o r k , q u a lita tiv e o b s e r v a t io n s a n d fo r m u la tio n o f
e m p ir ic a l c o r r e la tio n s fo r a p p lic a tio n t o th e d is p e r s io n m o d e l.
3.1.1 Experiment rationale

A r ig o r o u s a s s e s s m e n t o f t h e p h y s ic a l v a r ia b le s lik e ly t o b e e n c o u n t e r e d in t y p ic a l R A C
a p p lic a tio n s w a s m a d e t o d e v e lo p a t e s t p r o g r a m m e . T h e r e le v a n t c h a r a c t e r is tic s a s s o c ia t e d w ith
v a r i o u s R A C e q u i p m e n t w a s i n t e r p r e t e d in g e n e r i c t e r m s s o t h a t t h e r e s u l t s c o u l d b e e x t e n d e d t o
o th e r e q u ip m e n t. T o m in im is e th e n u m b e r o f t e s t s , c e r ta in v a r ia b le s w e r e f ix e d ( t y p ic a lly th o s e
t h a t c a n n o t b e c o n t r o l l e d b y e q u i p m e n t d e s i g n ) a c c o r d i n g t o a " '■ rea listic w o r s t c a s e ” , s i n c e i t i s
c o n v e n tio n a l in s a f e t y a s s e s s m e n t s n o t t o u n d e r e s tim a te th e s e v e r ity o f a h a z a r d . A w a ll-m o u n t e d
s p lit a ir c o n d it io n e r a ir -h a n d lin g u n it ( A H U ) w a s e m p lo y e d f o r th e m a j o r it y o f t h e e x p e r im e n t s
s in c e it is e a s i l y a d j u s t e d f o r t h e m a jo r ity 7o f c h a r a c t e r is t ic s o f in t e r e s t . I n a d d it io n , o th e r
e q u ip m e n t w a s e m p lo y e d fo r v e r if ic a t io n p u r p o s e s t o d e te r m in e w h e th e r in itia l o b s e r v a t io n s c o u ld
b e e x tr a p o la te d . T a b le 3 9 g iv e s a s u m m a r y o f th e v a r ia b le s e x p e c t e d t o h a v e a n in f lu e n c e o n th e
d is p e r s io n o f r e le a s e d r e fr ig e r a n t, a n d a n in d ic a tio n o f th e v a lu e s s e le c te d fo r th e t e s t s .
Refrigerant mass
M a s s o f r e f r ig e r a n t is c l o s e l y lin k e d t o t y p e , d e s ig n a n d c a p a c ity ' o f R A C e q u ip m e n t , a n d fr o m
t h e lite r a tu r e a la r g e r r e le a s e r a is e s c o n c e n t r a t io n s . E q u a t io n ( 3 0 8 ) r e p r e s e n ts t h e c o n v e n t io n a l
m e th o d fo r s p e c if y in g m a x im u m a llo w a b le m a s s o f H C w ith in a s p a c e , b e in g a fu n c tio n o f L F L .
N o r m a lly , sa fe ty s ta n d a r d s a d o p t a s a fe ty fa c to r o f fo u r ( (p - 0 .2 5 , A S H R A E 1 5 , 2 0 0 1 ) o r fiv e
( <p = 0 . 2 , E N 3 7 8 , 2 0 0 0 ).
mr,max=<P'LFL-VRm (3 0 8 )
E v a lu a tio n o f e q u a tio n ( 3 0 8 ) u s in g q> = 0 .2 , H C s ( w h o s e L F L r a n g e fr o m 0 . 0 3 8 k g .r r i3 t o 0 .0 4 5
k g .m " 3) g i v e 7 - 9 g m ° o f r o o m v o l u m e , k n o w n a s t h e P r a c t i c a l L i m i t . E N 3 7 8 ( 2 0 0 0 ) a l s o
s p e c i f i e s a n u p p e r l im it f o r H C s , w h ic h i s 1 .5 k g in p u b l i c l y o c c u p i e d s p a c e s ( e . g . , a s h o p ) , a n d
2 .5 k g i n p r i v a t e l y o c c u p i e d s p a c e s ( e . g . . a n o f f i c e ) . A c c o r d i n g l y , a r a n g e o f r e f r i g e r a n t m a s s e s
w ill b e e v a lu a te d u p t o a p p r o x im a te ly 2 .5 k g , b u t p r e d o m in a n tly 1 - 1 .5 k g s in c e t h is c o r r e s p o n d s
t o th e u p p e r lim it fo r th e t e s t c h a m b e r .
Table 39: Summary o f test parameters
Parameter Selected for tests
Equipment type 3 AHU, 2 refrigerators

Refrigerant release mass Up to 2.5 kg
Choice of gas/refrigerant Carbon dioxide
Room air movement variable flow rates from unit fan
Air discharge height Variable
Air discharge direction Horizontal, upwards, downwards
Post-ventilation time variable
Refrigerant release height Variable
Refrigerant release location Beneath fan. other points in room
Refrigerant release time Variable
Room size 135 m3
Floor surface Course carpet
Room tightness Fully sealed
Thermal internal convection None, people and PCs
Room air temperature 21°C
Room obstacles None, people and PCs
Release phase Vapour
Refrigerant release velocity < 0.1 m s'1
Refrigerant release direction Downward
R o o m s iz e
D u e t o p r a c t ic a l c o n s t r a in t s , o n l y a s in g le r o o m s i z e c o u ld b e t e s t e d . T h e lit e r a tu r e r e v ie w 7
h ig h lig h t e d s t u d ie s th a t s h o w e d th a t f o r a r e le a s e o f m a s s c o r r e s p o n d in g t o a f ix e d p r a c t ic a l lim it,
a la r g e r f lo o r a r e a t e n d e d t o p r o d u c e h ig h e r c o n c e n t r a t io n s ( e .g ., F le t c h e r a n d J o h n s o n , 1 9 8 6 ,
C C P S , 1 9 9 6 , K a ta o k a e t a l, 2 0 0 0 b ) . F le t c h e r a n d J o h n s o n a ls o c o m m e n t o n th e e f f e c t o f r o o m
s iz e w h e n v e n t ila t io n is u s e d , p o in t in g o u t th a t f o r a s p e c if ic a ir flo w 7 r a te , m ix i n g is l e s s e f f e c t iv e
f o r la r g e r r o o m s . In te r m s o f r o o m g e o m e tr y , s e v e r a l s tu d ie s ( e .g ., M a r s h a ll, 1 9 8 4 a n d C le a v e r e t
a l, 1 9 9 4 ) fo u n d th a t r o o m s h a p e a n d c e ilin g h e ig h t h a d n e g lig ib le in flu e n c e o n flo o r
c o n c e n t r a t io n s p a r t ic u la r ly in a s t ill e n v ir o n m e n t . T h u s , a r e la t iv e ly la r g e s p a c e w a s f a v o u r e d f o r
th e e x p e r im e n ts . A te s t c h a m b e r o f 6 .7 m * 6 . 7 m * 3 . 0 m in h e ig h t w a s c o n s t r u c t e d .
300
Floor surface
C o n v e n tio n a l d is p e r s io n m o d e ls fo r la r g e r e le a s e s o u t s id e h a v e a d e p e n d e n c e u p o n te r r a in
g e o m e tr y , w h e r e a n u n e v e n s u r fa c e te n d s t o p o o r e r m ix in g (C C P S , 1 9 9 6 , W o o d w a r d , 1 9 9 8 ).
A s s u m in g t h is a p p lie s e q u a ll y t o f lo o r s u r f a c e s in a c o n f in e d s p a c e , a r e la t iv e ly c o a r s e c a r p e t w a s
e m p lo y e d .
Room obstacles
F r o m t h e lit e r a tu r e , it is k n o w n th a t o b s t a c l e s d o n o t g e n e r a l ly in h ib it m ix i n g ( e .g . , L e a c h a n d
B lo o m f ie ld , 1 9 7 3 a n d B r itte r 1 9 8 8 ) , e x c e p t in r a r e o c c a s io n s w h e r e c h a n n e ls a r e c r e a te d o r w h e n
t h e r e is a b a r r ie r a t f lo o r le v e l. T h e n u m b e r o f p o s s i b l e t y p e s o f o b s t a c le s a n d c o m b in a t io n s
w o u ld m a s s i v e l y in c r e a s e t h e n u m b e r o f t e s t s . A c c o r d in g ly , it w a s d e c id e d n o t t o a c c o u n t f o r
o b s ta c le s .
Air temperature
T h e e f f e c t o f a ir t e m p e r a t u r e is ( a n d h u m id it y ) la r g e ly n e g l e c t e d in t h e lit e r a t u r e . H o w e v e r , s in c e
m e a s u r e m e n t s a r e g e n e r a l ly in t e r p r e te d i n t e r m s o f r e l a t iv e d e n s it y , o r b u o y a n c y o f t h e r e l e a s e ,
th e ir e f f e c t s a r e im p lic it. A h ig h e r a ir ( a n d r e le a s e ) t e m p e r a tu r e h a s t h e e f f e c t o f r e d u c in g
b u o y a n c y o f t h e r e f r ig e r a n t . G iv e n t h a t t h e a p p lic a t io n o f R A C e q u ip m e n t i s n o r m a ll y in
o c c u p ie d s p a c e s , a n o r m a l d e s ig n te m p e r a tu r e o f 2 1 ° C a n d 5 0 % r e la tiv e h u m id ity w a s u s e d .
Room air movement

T h e lit e r a tu r e id e n t if ie s r o o m a ir m o v e m e n t a s a m a j o r in f lu e n c e o n d is p e r s io n . A ir m o v e m e n t is
g e n e r a t e d b y m o v e m e n t o f th e g a s w it h in th e r o o m , a n d m a y a ls o b e p r o d u c e d b y in filtr a tio n ,
fo r c e d a ir f lo w a n d th e r m a l s o u r c e s (a d d r e s s e d b e lo w ) . S in c e R A C e q u ip m e n t o ft e n u t ilis e s a n
in t e g r a l f e n t o d is c h a r g e a ir fr o m t h e e v a p o r a t o r o r c o n d e n s e r in t o t h e r o o m , it w a s im p o r t a n t t o
u n d e r s ta n d it s e f f e c t o n d is p e r s io n . I n te g r a l f a n s w e r e t e s t e d u n d e r v a r y in g a ir f lo w r a te s,
d is c h a r g e v e lo c it ie s a n d d ir e c tio n s , a n d p o s it io n in g w ith in th e r o o m . E x te r n a l v e n tila tio n (in to a n d
e x tr a c te d fr o m th e r o o m ) w a s n o t e m p lo y e d .
A fu r th e r c o n s id e r a t io n is th e o p e r a tio n p e r io d fo r th e fa n . D u r in g o n - c y c le a fa n w ill n o r m a lly b e
r u n n in g a n d s w it c h e d o f f d u r in g o f f - c y c l e , a lt h o u g h a l o s s o f r e f r ig e r a n t lo w e r s s u c t io n a n d
d is c h a r g e p r e s s u r e s w h ic h m a y c a u s e t h e e v a p o r a t o r a n d /o r c o n d e n s e r fa n s t o b e te r m in a t e d .
W it h th is in m in d it is im p o r t a n t t o c o n s id e r th e w h e t h e r th e f a n n e e d s t o o p e r a t e c o n t in u o u s ly ,
w h ic h is d e c id e d a c c o r d in g t o t h e p o s s i b ilit y o f r e f r ig e r a n t le a k a g e d u r in g t h e p e r io d s th a t t h e
301
s y s t e m i f o f f . F u r t h e r , d u e t o th e r a p id c h a n g e in s y s t e m p r e s s u r e s , te m p e r a t u r e s a n d v ib r a t io n
f o ll o w i n g c o m p r e s s o r s ta r t-u p a n d s h u t - d o w n , t h e p o s s i b ilit y o f a le a k d e v e lo p in g a t t h e s e t im e s
is lik e ly t o in c r e a s e . G iv e n th a t a c a ta s tr o p h ic le a k m a y o c c u r a ft e r c o m p r e s s o r c e s s a t io n , th e r e is
a c a s e fo r p r o lo n g in g fa n o p e r a tio n f o llo w in g s h u td o w n ; th is c o n c e p t o f p r o tr a c te d a ir f lo w is
te r m e d t h e p o s t - v e n t ila t io n p e r io d . A s w it h c o m p r e s s o r te r m in a t io n , t h e e f f e c t o f c o m p r e s s o r
s ta r tin g m a y a l s o in it ia t e a le a k , b u t f o r t h is s it u a t io n w h e r e th e le a k m a y b e g in a t th e s a m e tim e
a s th e fa n , a ir flo w ’ w ill b e u n d e r d e v e lo p e d a n d t h e r e fo r e l e s s e f f e c t iv e d is p e r s io n . S im ila r ly ,
in itia tio n o f a ir flo w b y a p r e s s u r e s e n s o r o r r e fr ig e r a n t d e te c to r w o u ld h a v e th e s a m e e ffe c t.
T h e r e fo r e , t o s im u la te th is w o r s e c a s e a ir f lo w , t e s t s w e r e c o n d u c t e d b y in itia tin g a ir flo w
s im u lt a n e o u s ly w ith th e r e le a s e .
Room tightness
T h e t ig h t n e s s o f t h e r o o m in f lu e n c e s in f iltr a tio n , w h ic h in t r o d u c e s a ir flo w ’ th a t is d if f ic u lt t o
m e a s u r e . D e s p i t e t y p ic a l in f iltr a tio n r a te s b e in g s m a ll a ir flo w r a te s o f le s s th a n 0 .5 h '1 h a v e b e e n
s h o w n t o in f lu e n c e f lo o r c o n c e n tr a tio n (L e a c h a n d B lo o m fie ld , 1 9 7 3 , F le tc h e r a n d J o h n s o n , 1 9 8 6 ,
M a r s h a ll, 1 9 8 4 ). T h e r e fo r e , a g a s -tig h t s e a le d r o o m w a s c o n s id e r e d e s s e n tia l t o a v o id e r r o n e o u s
in te r p r e ta tio n o f m e a s u r e m e n ts . S e a lin g th e r o o m a ls o p r e v e n ts m ig r a tio n o f r e fr ig e r a n t th r o u g h
c r a c k s o r p o r o s ity o f th e w a lls .
Thermal convection
C le a v e r e t a l ( 1 9 9 4 ) s ta te d th a t th e r m a lly in d u c e d c o n v e c t io n fr o m w a r m s u r fa c e s c o u ld p r o d u c e
s ig n ific a n t m ix in g , w h ic h im p lie s th a t h e a t s o u r c e s s h o u ld b e r e m o v e d fr o m th e r o o m . H e a t
t r a n s f e r t h r o u g h w a l l s c r e a t e s a d if f e r e n t ia l te m p e r a t u r e a c r o s s t h e a ir la y e r a t t h e w a l l , a n d t h u s
c o n v e c t io n . T o a v o id t h is , t h e t e s t c h a m b e r w a s lo c a t e d w it h in a la r g e r s p a c e w h ic h w a s
m a in ta in e d a t th e s a m e te m p e r a tu r e . I n s tr u m e n ta tio n w ith in th e r o o m w a s n o t e x p e c t e d t o
g e n e r a te a n y n o t ic e a b le h e a t o u tp u t.
Height of airflow source (fan)

F r o m M a r s h a ll ( 1 9 8 4 ) a n d F le tc h e r a n d J o h n s o n ( 1 9 8 6 ) , th e h e ig h t o f th e v e n tila tio n in le t a n d
o u tle t a r e c r itic a l t o m ix in g , p a r tic u la r ly w ith r e s p e c t t o th e p o s it io n o f th e r e le a s e s o u r c e . S in c e
t h e f a n h e ig h t v a r ie s w id e ly w it h e q u ip m e n t t y p e , it is c o n s id e r e d a n im p o r t a n t v a r ia b le .
A p p r o p r ia te v a lu e s w e r e s e le c te d a c c o r d in g t o t y p ic a l in s ta lla tio n h e ig h ts fo r e q u ip m e n t, c o v e r in g
flo o r t o c e ilin g m o u n te d .
302
Airflow discharge direction
W h ils t m o s t c o n d e n s in g u n its d is c h a r g e a ir in a s in g le d ir e c tio n , m a n y A H U h a v e a m u lti
d ir e c tio n a l d is c h a r g e v e n t. A c c o r d in g t o M a r s h a ll ( 1 9 8 4 ) a n d F le tc h e r a n d J o h n s o n ( 1 9 8 6 ) ,
a ir f lo w d ir e c tio n r e la tiv e t o th e d ir e c tio n o f th e r e le a s e is a s ig n ific a n t fa c to r in th e a m o u n t o f
m ix i n g . T h e r e f o r e u p w a r d , h o r iz o n t a l o r d o w n w a r d a ir d is c h a r g e s fr o m t h e A H U w e r e a s s e s s e d .
Gas/refrigerant
S e v e r a l H C s a r e lik e ly t o b e u s e d in a ir - c o n d it io n in g e q u ip m e n t: R 2 9 0 , R 6 0 0 a a n d R 1 2 7 0 , a n d
m ix tu r e s th e r e o f. S in c e th e m o d ific a tio n o f te s t f a c ilit ie s a n d p u r c h a s e o f s p e c ia lis t
in s tr u m e n ta tio n t o m it ig a te th e p o te n tia l fo r ig n itio n w h e n d e a lin g w ith fla m m a b le g a s e s im p o s e s
e x c e s s iv e c o s t , a n o n -fla m m a b le m im ic w a s u s e d . C a r b o n d io x id e w a s c h o s e n b a s e d o n s im ila r
d e n s ity t o R 2 9 0 a n d R 1 2 7 0 a t r o o m c o n d itio n s , a n d th e la r g e n u m b e r o f s tu d ie s th a t s u c c e s s f u lly
e m p lo y e d it a s a n a lt e r n a t iv e t o L P G in d is p e r s io n t e s t s ( e .g . , L e a c h a n d B lo o m f i e ld , 1 9 7 3 ,
F le tc h e r a n d J o h n s o n , 1 9 8 6 , C le a v e r e t a l, 1 9 9 4 , K a ta o k a e t a l, 2 0 0 0 a , 2 0 0 0 b ) .
Release rate
F r o m v a r io u s s t u d ie s ( F le t c h e r a n d J o h n s o n , 1 9 8 6 , C le a v e r e t a l, 1 9 9 4 ) , it is k n o w n t h a t t h a t a
h ig h e r r e le a s e f l o w r a te p r o d u c e s h ig h e r c o n c e n tr a tio n s . F o r a f in it e c h a r g e , t h e m a s s flo w 7 r a te o f
t h e r e fr ig e r a n t ( a s s u m e d to b e c o n s ta n t) d ic ta te s th e r e le a s e tim e , w h ic h im p lie s th a t th e s h o r te s t
t im e fo r t h e t o t a l c h a r g e t o b e le a k e d w ill p r o d u c e h ig h e r c o n c e n t r a t io n s . T h e t h e o r e t ic a l
m in im u m r e le a s e tim e , t r m in , o c c u r r in g fr o m a fu ll-b o r e p ip e s h e e r w h ils t th e s y s t e m is o p e r a tin g
c a n b e e s t im a te d fr o m th e m a s s f lo w r a te a n d r e fr ig e r a n t c h a r g e ( mr ) , (e q u a tio n 3 0 9 )
m, ■AH
^ (309)
~0
w here AH is r e fr ig e r a n t e n th a lp y d if fe r e n c e a c r o s s th e e v a p o r a to r a n d O is r e fr ig e r a tin g
c a p a c i t y 7. F r o m r e v i e w i n g m a n u f a c t u r e r d a t a , t y p i c a l H C c h a r g e s i z e s e q u a t e t o 0 . 1 - 0 .2 k g k W '1
o f e v a p o r a to r c a p a c ity . E v a lu a tin g e q u a tio n ( 3 0 9 ) fo r s e v e r a l H C s fo u n d trmin t o b e around one
m in u te . H o w e v e r , u n d e r r e a l c o n d it io n s a lo s s o f r e fr ig e r a n t s t o p s th e c o m p r e s s o r , th u s s lo w in g
r e le a s e r a te .
A n o th e r a p p r o a c h is t o c o n s id e r th e r e fr ig e r a n t f lo w th r o u g h a le a k h o le o f a g iv e n s iz e . F o r
e x a m p le , th e r e p o r ts o f B lo m - B r u g g e m a n e t a l ( 1 9 9 6 a , 1 9 9 6 b ) u s e th e s te a d y f lo w e q u a t io n t o
c a lc u la t e m a s s f l o w o f r e f r ig e r a n t t h r o u g h a f u ll b o r e p ip e r u p t u r e , b u t r e s u lt s in v e r y r a p id
r e le a s e t im e s , f a r s h o r te r th a n e q u a t io n ( 3 0 9 ) . O th e r lit e r a tu r e r e p o r t s o n t e s t s t o d e te r m in e
303
c a t a s t r o p h ic le a k t im e , in c lu d in g K a t a o k a ( 1 9 9 9 ) w h o g a v e a t im e o f 4 m in u te s fr o m a n a ir
c o n d itio n e r a n d B la c k w e ll a t a l ( 2 0 0 4 ) m e a s u r e d 1 9 6 s u s in g 0 .3 9 k g o f C O ? a n d 1 4 6 s fo r 0 .4 2
k g o f R 1 5 2 a f o r a c a r a ir c o n d it io n in g s y s t e m . A m o d e l fo r c a t a s t r o p h ic r e le a s e s fr o m c h e m ic a l
p la n t w a s r e p o r t e d b y W o o d w a r d ( 1 9 9 9 ) th a t t a k e s a c c o u n t o f in t e r n a l f r ic t io n a l lo s s e s a s th e
flu id f lo w th r o u g h p ip e w o r k a n d fit tin g s . T h is w a s te s te d b y u s in g in p u ts a p p r o p r ia te fo r a
r e fr ig e r a tin g c ir c u it b u t w a s fo u n d t o r e tu r n u n r e a lis tic r e s u lts .
F o r th is w o r k a n u m b e r o f c a ta s tr o p h ic le a k a g e e x p e r im e n ts w e r e m a d e ( P R 1 0 0 3 7 4 , 2 0 0 0 ) a n d
f o u n d th a t t h e r e f r ig e r a n t r e le a s e d f r o m a n a ir c o n d it io n e r c e a s e d a f t e r a b o u t 3 m in u te s ,
a p p r o x im a te ly c o n s is t e n t w ith th e r e s u lts fr o m K a ta o k a e t a l ( 1 9 9 9 ) a n d B la c k w e ll a t a l ( 2 0 0 4 ) .
T h e r e fo r e , a m in im u m r e le a s e tim e o f 2 1 0 s w a s c o n s id e r e d r e a s o n a b le fo r th e R A C e q u ip m e n t
p r e s e n t ly u n d e r c o n s id e r a t io n . W h il s t t h is r e le a s e tim e r e p r e s e n ts a c a t a s t r o p h ic le a k , p r o lo n g e d
r e le a s e s a ls o o c c u r . A lth o u g h th e e f f e c t o f a lo n g e r r e le a s e tim e fo r a g iv e n m a s s is a lo w e r flo o r
c o n c e n tr a tio n , a r e le a s e d u r a t io n th a t e x c e e d s t h e p o s t - v e n t ila t io n p e r io d h a s th e p o t e n t ia l t o
p r o d u c e h ig h e r c o n c e n t r a t io n s . I n t h is s it u a t io n , it w a s im p o r t a n t t o u n d e r s t a n d t h e in f lu e n c e o f
p o s t-v e n t ila tio n a ir f lo w o n s u b s e q u e n t m ix in g . N o t e th a t a lth o u g h le a k tim e c a n n o t b e c o n tr o lle d
b y d e s ig n , it w a s t e s t e d o v e r a r a n g e o f v a lu e s b e c a u s e th e w o r s e c a s e c h a n g e d w it h d if f e r e n t
p o s t-v e n t ila tio n p e r io d s .
R e le a s e v e lo c ity
S tu d ie s b y B a in e s a n d T u r n e r ( 1 9 6 9 ) , M a r s h a ll ( 1 9 8 4 ) a n d C le a v e r e t a l ( 1 9 9 4 ) c o n s id e r e d th e
e f f e c t o f r e l e a s e v e lo c it y . In a ll, it w a s s h o w n t h a t a h ig h e r v e l o c i t y p r o d u c e d im p r o v e d m ix in g ,
p a r tic u la r ly w h e r e th e r e le a s e p o in t is c lo s e t o th e flo o r . G e n e r a lly , a le a k fr o m a r e fr ig e r a tio n
c ir c u it w ill b e a t h ig h v e lo c it y d u e t o t h e la r g e p r e s s u r e d if f e r e n c e . H o w e v e r , in s o m e c a s e s th e
le a k m a y im p in g e o n th e in s id e s u r f a c e o f a n e n c lo s u r e c a u s in g t h e r e le a s e in to t h e r o o m t o h a v e
n e g lig ib le m o m e n tu m . E x a m p le s o f th is in c lu d e r e le a s e s in s id e a r e fr ig e r a te d c o m p a r tm e n t o r
w it h in A H U c a s in g . D e s p i t e th e o b s e r v a t io n o f M a r s h a ll th a t t h e e f f e c t o f r e le a s e v e lo c ity ' is
in h ib it e d w h e n th e r o o m is v e n t ila t e d , it w a s c o n s id e r e d im p o r t a n t t o e m p lo y lo w v e lo c ity ' r e le a s e s
t h r o u g h o u t t h e t e s t s . F r o m t h e l i t e r a t u r e , v e l o c i t y i n f l u e n c e s a r e m i n i m a l b e l o w 0 . 1 m s ' 1, s o a
d iffu s e r d e v ic e w a s e m p lo y e d t o e n s u r e th is w a s n o t e x c e e d e d .
T w o - p h a s e r e l e a s e s a l s o o c c u r f r o m r e f r ig e r a n t c ir c u it s , b u t in lig h t o f t h e d i s c u s s i o n o f liq u id
r e l e a s e s in t h e lit e r a tu r e r e v ie w , a t w o - p h a s e r e le a s e is e x p e c t e d t o p r o d u c e m o r e f a v o u r a b le
m ix in g s o v a p o u r o n ly r e le a s e s w e r e c o n s id e r e d . Q u a n t if ic a t io n o f t h e m o d e ls in W o o d w a r d
304
( 1 9 9 9 ) s h o w e d th a t f o r a liq u id R 2 9 0 r e le a s e in to t h e t e s t c h a m b e r w o u ld h a v e a n e x p a n s io n r a te
e q u iv a l e n t t o 2 - 3 r o o m v o lu m e c h a n g e s p e r h o u r , s u f f i c ie n t f o r r a p id d is p e r s io n o f t h e g a s .
Release direction
T h e e f f e c t o f r e le a s e o r ie n ta tio n w a s r e p o r te d b y B a in e s a n d T u r n e r ( 1 9 6 9 ) . B a in e s e t a l ( 1 9 9 0 )
a n d B a r n e t t ( 1 9 9 1 ) , w h ic h in d ic a t e d th a t a d o w n w a r d d ir e c t io n le a d s t o h ig h e r f lo o r
c o n c e n tr a tio n s . In th e c a s e o f v e r y lo w v e lo c it y r e le a s e s th o u g h , th e e f f e c t o f o r ie n ta tio n b e c o m e s
m in o r s in c e th e m o m e n tu m o f th e p lu m e is s m a ll c o m p a r e d to b u o y a n c y fo r c e s . F o r th e p u r p o s e s
o f th e t e s ts , o n ly a d o w n w a r d r e le a s e w a s u s e d .
Release height
L ite r a tu r e c o n s is t e n t ly r e p o r te d th a t t h e m e a n c o n c e n t r a t io n w it h in a p lu m e r e d u c e s w it h d is t a n c e
fr o m th e s o u r c e , w h ic h is r e fle c te d b y f lo o r c o n c e n tr a tio n s . O b s e r v a t io n s o f R A C e q u ip m e n t
r e v e a l th a t le a k s c o u ld o c c u r fr o m a v a r ie ty o f h e ig h ts , d e m a n d in g te s tin g o v e r a s im ila r r a n g e .
H o w e v e r , s o m e l e a k p o s i t i o n s c o u l d b e i g n o r e d b e c a u s e t h e y a r e l i k e l y t o r e s u l t in t h o r o u g h
m i x i n g a n y w a y , a s s e e n in t h e m e a s u r e m e n t s o f G o e t z l e r e t a l ( 1 9 9 8 ) . S u c h p o s i t i o n s i n c l u d e d t h e
c a s e o f a r e l e a s e o c c u r r i n g f r o m w i t h i n t h e a i r f l o w d u c t o r d i r e c t l y in t h e a i r d i s c h a r g e , a n d
r e l e a s e s f r o m a b o v e t h e a ir d is c h a r g e .
Release location
M o s t s tu d ie s th a t a d d r e s s e d v e n tila tio n p r e s e n t r e s u lts b a s e d o n r e le a s e s o v e r a r a n g e o f h e ig h ts
b u t la r g e ly ig n o r e la te r a l p o s itio n in g . In th e c a s e o f a s p lit - s y s t e m in s ta lla tio n , p ip in g is o ft e n
fitte d a lo n g w a lls a n d c e ilin g s t o th e e v a p o r a to r , th u s th e p o t e n tia l fo r a r e le a s e a t p o s it io n s o th e r
t h a n d ir e c t ly b e lo w t h e fa n . A c c o r d in g ly , it w a s n e c e s s a r y t o a s s e s s r e le a s e s o u r c e s fr o m
p o s it io n s a r o u n d th e r o o m a w a y fr o m th e s o u r c e o f a ir flo w .
3.1.2 Apparatus set-up and experimental procedure

A g e n e r a l s c h e m e o f h o w th e te s t c h a m b e r w a s s e t u p , a c c o r d in g t o th e d is c u s s io n o f v a r ia b le s , is
i l l u s t r a t e d i n F i g u r e 1 0 2 . 24 T h e C 0 2 a n d a s s o c i a t e d c o n t r o l e q u i p m e n t w a s l o c a t e d o u t s i d e t h e
r o o m , a s w a s th e C 0 2 a n a ly s e r . A fte r w a r m in g a n d m o d u la tin g th e C 0 2 a c c o r d in g t o th e flo w
m e te r it w a s p a s s e d th r o u g h a tu b e t o a d if f u s e r lo c a t e d w it h in t h e r o o m . A b o v e th e d if f u s e r w a s
t h e A H U w it h t h e r e s p e c t iv e a ir d is c h a r g e a n d in le t g r ille s , a n d t h e f a n w a s in itia te d a n d s t o p p e d
u s i n g a n e x t e r n a l l y l o c a t e d s w i t c h . S a m p l i n g p o i n t s d r e w t h e C 0 2/ a i r m i x t u r e f r o m v a r i o u s p o i n t s
24 The experiments were carried out within environmental chambers at BRE laboratories, Watford, UK.
305
in the room and the resulting concentrations were determined with the C 0 2 analyser. Temperature
and velocity measurements were also made at various locations within the chamber.
The test chamber (6.7 m * 6.7 m * 3.0 m in height) provided an overall empty volume o f 135 m3.
The additional room contents such as the air-handier, tri-pods and instrumentation did not
contribute to the depletion o f room volume by any more than 0.5 m3 and the impact on room
volume was considered negligible. To ensure the room was airtight, it was fully sealed internally
with polyethylene sheeting. A decay test was conducted on the room each day to check tightness
over the duration o f a normal test period, w hich exhibited no noticeable decay in concentration.
regulator.
flow
m eter and sam pling
heater air discharge
lines
air return
diffuser
CO, cylinders
Figure 102: Schematic arrangement o f the test chamber apparatus
High purity C 0 2 w as used to avoid practical problems associated with the presence o f moisture
and other impurities as well as property calculations. To ensure sufficient gas be released within
the specified time, four cylinders w ere placed within a warm w ater bath. The outlet o f the
cylinder was connected to a heated regulator to prevent freezing and to ensure a discharge
temperature close to ambient. The outlet o f the regulator was piped to a pressure transducer and
an analogue flow meter and then to the diffuser located within the room. The regulator was
controlled manually so that a constant flow?rate was indicated on the flow meter. A diffuser was
used to provide a mean gas exit velocity o f betw een 0.05 m s'1 and 0 .1 m s'1 across the
anticipated range o f mass flow rates. The construction o f this diffuser was primarily o f a plastic
funnel with 10 mm inlet diameter and 190 mm outlet diameter and 200 mm long. The funnel was
stuffed with insulating fibre to even radial velocity and course gauze used to hold in the fibre. The
smoke used for visualisation tests w as generated using a “Smokefactory Sky Walker’’ fog
generator.
306
T h e R A C e q u ip m e n t c o m p r is e d th r e e A H U s a n d t w o r e f r ig e r a te d c a b in e t s , a s d e ta ile d in T a b le
4 0 . T h e I m p a c t 1 0 0 u n it w a s c h o s e n f o r t h e m a j o r it y o f t h e t e s t s f o r s e v e r a l r e a s o n s . A p a r t fr o m
a n A H U b e in g m o r e v e r s a t ile , th e I m p a c t 1 0 0 h a d a m a n u a l u n id ir e c t io n a l a ir d is c h a r g e g r ille th a t
c o u ld b e e a s il y r e d ir e c te d . In a d d itio n , t h e v e lo c it y in to t h e u n it w a s c o n s id e r a b ly lo w e r th a n th e
o t h e r w a ll m o u n t e d A H U , w h ic h h e lp e d m it ig a t e t h e in f lu e n c e o f t u r b u le n c e f r o m t h e r e tu r n a ir .
Table 40: Information on RAC equipment

Open front Refrigerated
Wall mounted Wall mounted Ceiling AHU
Unit type display storage
AHU #1 AHU #2 cassette #3
cabinet #4 cabinet #5
Name Impact 100 PKH 3 PLH 4 ZM Gastronorm
Dimension of
1.4 m x 0.5 m 1.4 m x 0.3 m 1.5 m x 0.6 m 0.6 m x 1.3 m 2.1 m x 0.7 m
air outlet
Nominal
9.9 kW T 7.7 kW 1 9.7 k W 1 1.7 kW 0.4 kW
Capacity
Nominal
0.96 k g 1 1.30 kg : 1.68 kg* 0.80 kg 0.35 kg
Charge
Test regime Extensive Selected Selected Selected Selected
+ At defined standard conditions.1 With standard 7m interconnecting pipe.
Instrumentation
A l i s t o f i n s t r u m e n t a t i o n i s g i v e n in T a b l e 4 1 . C l e a v e r ( 1 9 9 9 ) s u g g e s t s a m i n i m u m o f 6 s a m p l e
p o in ts w ith 2 m e a s u r in g c o n tin u o u s ly . It s h o u ld b e n o te d th a t th e r e is a tim e la p s e b e tw e e n a c tu a l
e v e n ts a n d g a s a n a ly s e r m e a s u r e m e n ts d u e t o th e tim e ta k e n to d r a w th e a ir /g a s th r o u g h th e
s a m p l e t u b e s . P o s it i o n i n g o f s a m p le p o in t s a n d t h e t h e r m is t o r s / a n e m o m e t e r s w a s in a c c o r d a n c e
w i t h t h e c o o r d i n a t e s in T a b le 5 7 o r ie n t a t e d a c c o r d i n g t o F i g u r e 1 8 6 ( A p p e n d i x 2 ) .
C o n c e n t r a t io n d a t a w a s o b t a in e d in t h e f o r m o f a v o l t a g e o u t p u t f r o m g a s a n a l y s e r s . C a lib r a t io n
r e q u ir e d v o lt a g e s th a t c o r r e s p o n d e d t o k n o w n c o n c e n t r a t io n s a c r o s s t h e n o r m a l o p e r a tin g r a n g e ;
t h e s e w e r e o b t a i n e d w i t h c a l i b r a t i o n g a s e s a t s t a n d a r d a t m o s p h e r i c c o n c e n t r a t i o n ( 0 . 0 0 0 7 k g m ' 3) ,
0 . 0 0 9 k g m '3 a n d 0 . 0 1 8 k g m '3.O u t p u t v o l t a g e s w e r e c o n v e r t e d t o a c t u a l m e a s u r e d c o n c e n t r a t io n
( Cactm) u s in g e q u a tio n ( 3 1 0 ) .
Cactm= o,-P D + a2 (3 1 0 )
w here a{ a n d a2 a r e c o n s ta n ts (o b ta in e d fr o m a d a ily c a lib r a tio n r o u tin e ) a n d PD is th e o u tp u t
v o lta g e . G iv e n th a t a b a c k g r o u n d c o n c e n tr a tio n o f C 0 2 e x is t s , a n “ a b s o lu t e ” c o n c e n tr a tio n w a s
307
c a lc u la t e d b y s u b tr a c tin g t h e a c t u a l m e a s u r e d c o n c e n t r a t io n m e a s u r e d d u r in g t h e t e s t s fr o m th e
m e a s u r e d c o n c e n t r a t io n p r io r t o t h e b e g in n in g o f t h e t e s t , C actm {0). T h e a b s o lu t e c o n c e n tr a tio n ,
C , w a s o b ta in e d fr o m e q u a tio n ( 3 1 1 )
^~actm C actm(0) (3 1 1 )
w h e r e C artm(0-, w a s a l w a y s a b o u t 0 . 0 0 0 7 k g m"3.
Table 41: Information on instrumentation used
Measurement Instrumentation details

Concentration 2 x Leybold Binos 1 infra-red analysers. 0 - 0.5 % (by volume) 5 * ADC infra-red
analysers. 2 with a 0% to 3% range, and three with a 0 % to 5 % range (one fitted to
a multi-point selector unit to give four intermittent sampling points). Sampling point
lines leading to analysers. The concentration is sampled at an interval of 5 seconds.
Flow rate Row meter to measure gas (CO?) flow rate for simulated refrigerant release
(analogue)
Cylinder weight. Salter Digital Scale for measuring gas release from cy linder.
Air velocity and 22 DANTEC combined spherical omni-directional hot-film anemometer and
temperature. thermistor sensors connected to a DANTEC 54N10 multi-channel flow analyser for
measurement of air speed and dry bulb air temperature. Both velocity and
temperature measurements are recorded every 200 seconds, although velocity is
monitored continuously over an integrated time period to pro\ide the relative
turbulence intensity.
Data acquisition Grant Instruments (Cambridge) Ltd Squirrel Meter/Logger 1000
Procedure
T h e c h o ic e o f te s t c o n d itio n s w a s la r g e ly d e te r m in e d o n a n e v o lv in g b a s is , w h e r e r e s u lts w e r e
a n a ly s e d f o ll o w i n g e a c h t e s t t o h e lp d e c id e w h e th e r fu r th e r in v e s t ig a t io n o f e a c h v a r ia b le w a s
n e c e s s a r y . T h is a p p r o a c h w a s e m p lo y e d s o th a t u s e o f r e s o u r c e s c o u ld b e o p t im is e d . T h e in itia l
s t e p w a s t o o b t a in a v i s u a l a p p r e c ia t io n o f t h e b e h a v io u r o f a r e l e a s e , w -h ic h w a s a c h ie v e d b y
film in g s m o k e te s ts . F o llo w in g th is c a r b o n d io x id e r e le a s e s w e r e m a d e a n d c o n c e n tr a tio n s
m easu red .
D u r in g th e fir s t s e t o f v is u a lis a tio n te s ts th e s m o k e w a s r e le a s e d fr o m v a r io u s p o in ts a b o u t th e
u n it a n d t h e r o o m . T h is p r o v id e d a b a s ic a p p r e c ia t io n o f h o w t h e A H U m o v e d a ir w it h in t h e
r o o m . T h e s e c o n d s e t o f s m o k e t e s t s w a s a im e d t o w a r d s v i s u a l i s i n g t h e e f f e c t o f t h e a ir flo w ' o n
t h e r e f r ig e r a n t r e le a s e . In o r d e r t o c o u n t e r a c t t h e n e u tr a l b u o y a n c y o f t h e s m o k e , it w a s m ix e d
w it h a f lo w 7o f c a r b o n d io x id e w ith in a d u c t p r io r t o its r e le a s e s o a s t o g iv e th e s m o k e a n
a r t i f i c i a l l y i n c r e a s e d b u l k d e n s i t y 7. O b s e r v a t i o n s w e r e u s e d t o d e t e r m i n e t h e b e s t p o s i t i o n i n g f o r
308
s a m p lin g p o in ts s o th a t th e h ig h e s t c o n c e n tr a tio n s w e r e m e a s u r e d . T h e c a r b o n d io x id e r e le a s e
t e s ts a ll f o llo w e d th e p r o c e d u r e b e lo w .
i. Initiate ventilation to remove excess carbon dioxide from previous test, so that
concentration was at the background level.
ii. Position air handler and gas release diffuser according to test schedule.
iii. Check positioning o f sampling point lines and anemometers.
iv. W eigh C 0 2 cylinder and note mass.
V. Switch o ff room ventilation system and seal dampers.
vi. Initiate recording o f data loggers.
vii. Wait several minutes for room air movement to cease.
viii. Switch on air handler fan from remote position.
ix. Wait several minutes for room airflow to reach steady state.
X. Open gas-regulating valve to obtain prescribed flow rate.
xi. After prescribed release duration terminate gas flow.
xii. After prescribed post-ventilation period switch o ff air handler fan.
xiii. Monitor room concentration until near-steady state is achieved.
xiv. 'Clean' room o f C 0 2 by initiating room ventilation system.
XV. W eigh CO: cylinder and note mass.
F o r th e c o n c e n tr a tio n m e a s u r e m e n ts , c o n s is t e n c y o f r e s u lts w e r e a c h ie v e d b y th e f o llo w in g
p r o c e d u r e c o n d u c t e d e a c h d a y : v is u a l c h e c k fo r r o o m s e a lin g , c a lib r a te c a r b o n d io x id e a n a ly s e r s ,
a n d c h e c k c a r b o n d io x id e a n a ly s e r d r y e r s a n d r e p la c e c r y s t a ls i f n e c e s s a r y . In a d d itio n t o th e s e , a
d e c a y t e s t o n t h e r o o m w a s c o n d u c t e d d a ily t o c h e c k f o r r o o m a ir t ig h t n e s s .
Test schedule
B a s e d o n th e e a r lie r d is c u s s io n , s p e c if ic c o m b in a tio n s o f v a r ia b le s w e r e c h o s e n . I n itia lly , a
r e fe r e n c e t e s t w a s e s t a b lis h e d s o th a t r e p e a ta b ility c o u ld b e c h e c k e d a n d t o a c t a s a b a s is fo r
c o m p a r is o n o f r e s u lts . T h e v a r ia b le s s e le c te d fo r th e r e fe r e n c e c o n d itio n s a r e a s fo llo w s : w a ll
m o u n te d I m p a c t 1 0 0 A H U , w ith h o r iz o n ta lly d is c h a r g e d a ir f lo w a t 0 .2 6 ir f s '1 fr o m a h e ig h t o f
2 .3 5 m a n d a p p r o x im a te ly 1 k g o f C 0 2 r e le a s e d fr o m 1 .9 5 m d ir e c tly b e lo w th e A H U o v e r 2 1 0 s.
V a r i a b l e s w e r e a d j u s t e d t h r o u g h o u t t h e t e s t in g , a n d t h e f in a l s c h e d u le lis t e d in T a b l e 4 2 ; f u ll
d e t a ils a r e p r o v id e d in T a b le 5 8 .
309
Table 42: General test schedule
Purpose (variable) Test no.

Smoke visualisation (reference, with airflow, airflow' + CO:) 1.2. 3
Decay/tightness (still) 4, 5. 7
Decay/tightness (with air movement) 6 .8
No air movement 9
Reference test 10. 24, 30
Air flow rate 11. 12, 26. 27, 28
Post-ventilation period and release time 13, 25, 29
Air discharge direction 14, 15
Release mass 16, 17
Horizontal release location 18. 31
Fan and release height 19. 20, 21. 22, 23
Change to AHU #2 (reference, airflow' rate, release height) 32. 33, 34
Change to AHU #3 (reference, reference, release height) 35. 36, 37
Change to cabinet #1 (reference, release mass, release mass) 38. 39, 40
Change to cabinet #2 (reference, release height) 41.42
Thermal manikins and PCs 43
Interpretation of data
M e a s u r e d d a ta fr o m e a c h te s t w a s c o n v e r t e d in to a u s a b le fo r m a t s o it c o u ld b e e m p lo y e d to
id e n tify tr e n d s a s s o c ia te d w it h th e s e le c te d v a r ia b le s . T h r e e c o n c e n tr a tio n s w e r e e x tr a c t e d fr o m
th e m e a s u r e m e n ts a s s h o w n b e lo w , a n d w e r e tr e a te d a c c o r d in g t o th e e x p la n a tio n s .
Mean room concentration. In o r d e r t o e s t im a te th e m a s s o f r e fr ig e r a n t w ith in th e r o o m fr o m th e
r e le a s e , th e m e a n r o o m c o n c e n tr a tio n w a s m u ltip lie d b y r o o m v o lu m e . A t th e e n d o f e a c h te s t,
a lt h o u g h t h e r o o m a ir w a s m ix e d , f in a l a n a ly s is o f t h e d a t a s h o w e d t h a t f u lly h o m o g e n o u s
d is tr ib u tio n o f C 0 2 d id n o t a lw a y s o c c u r , s o t h e m e a n r o o m c o n c e n t r a t io n ( CR m ) w a s c a lc u la te d
u s in g v o lu m e w e ig h tin g fo r e a c h s a m p lin g lo c a tio n (e q u a tio n 3 1 2 ) .
f 10
Z c rK ' VRm (3 1 2 )
V »=1
w here Vi is th e r e p r e s e n ta tiv e v o lu m e fo r a s p e c if ic s a m p lin g p o in t, a s T a b le 4 3 . R o o m
r e fr ig e r a n t m a s s w a s th e n c a lc u la te d u s in g e q u a tio n ( 3 1 3 ) .
m r,Rm = ' ^R m (3 1 3 )
310
A m a s s b a la n c e w a s u s e d t o c h e c k th e a m o u n t r e le a s e d fr o m th e c y lin d e r a g a in s t th e a m o u n t
p r e s e n t in th e r o o m . T h e m a s s c a lc u la te d fr o m e q u a tio n ( 3 1 3 ) w a s c o m p a r e d a g a in s t th e
d if f e r e n c e in c y lin d e r m a s s b e f o r e a n d a ft e r t h e r e le a s e ( s e e T a b le 5 8 ) , w h ic h s h o w e d g o o d
a g r e e m e n t , w it h in ± 5 % . F o r t h e p u r p o s e s o f in te r p r e ta tio n o f t h e r e s u lt s , th e m a s s b a s e d o n r o o m
c o n c e n t r a t io n w a s e m p lo y e d in s t e a d o f b y c y lin d e r w e ig h t , s in c e it c o r r e s p o n d s d ir e c t ly t o t h e
m e a s u r e d c o n c e n t r a t io n s , w h ic h a r e o f in t e r e s t in t h is s t u d y .
Table 43: Sample point location and representative volume

Ratio to define
A representative
Sample volume for
Location (key) Release Release
point sample point
at at room
centre wall Vi ivRm
C5 Floor level, back (0 m. back) 0.129 0.135 0.019
Cio Floor level, beneath unit (0 m. front) 0.129 0.139 0.019

c6 Floor level, middle of room (0 in. mid) 0.337 0.135 0.019
c8 Ceiling level, middle of room (3 m, mid) - - 0.417
c, Floor lev el, left of unit (0 m, LH front) 0.138 0.061 0.038
C: Floor lev el, left at mid room (0 m, LH mid) 0.129 0.350 0.038
c3 Floor lev el, left at back room (0 m. LH back) 0.138 0.180 0.038
C7 1 m above floor, middle (1 m, mid) - - 0.140
Q, 1 m above floor, beneath unit (1 m, front) - - 0.140
c4 1 m above floor, back of room (1 m. back) - - 0.140
Maximum floor concentration. O n e o f th e k e y a r e a s o f in t e r e s t in th is p a r t o f t h e s t u d y is th e
m a x im u m c o n c e n tr a tio n ( Cf m a x ) r e s u ltin g fr o m a r e le a s e . Cf max i s th e h ig h e s t c o n c e n tr a tio n
m e a s u r e d ( t y p ic a lly a t f lo o r le v e l) fo r t h e d u r a t io n o f e a c h te s t , a n d is o b t a in e d fr o m c h e c k in g
e a c h c o m p l e t e s e t o f s a m p l e d d a t a . T h e e x c e p t i o n i s d i r e c t l y w i t h i n o r b e l o w t h e r e l e a s e , s i n c e it
r e p r e s e n ts th e c o n c e n tr a tio n o f th e p lu m e o n ly . C o n c e n tr a tio n s fr o m th is s a m p lin g p o in t w e r e
o n ly in c lu d e d in th e a n a ly s is fr o m th e tim e a ft e r th e r e le a s e h a d s to p p e d . In a ll c a s e s , t h e r e s u lts
sh o w e d th a t C fmax o c c u r r e d w ith in a f e w s e c o n d s o f th e tim e o f c e s s a t io n o f th e r e le a s e .
Maximum mean floor concentration. T h e r e a r e p o s s ib le p r o b le m s a s s o c ia te d w ith th e a p p r o a c h
o f u s in g C fmax a s a r e p r e s e n ta tiv e c o n c e n tr a tio n . G iv e n th e lim ite d n u m b e r o f s a m p lin g p o in ts
a n d t h e e r r a tic n a t u r e o f c h a n g e s in c o n c e n t r a t io n o v e r t im e - w h ic h le a d t o h ig h lo c a lis e d
c o n c e n t r a t io n s - e r r o n e o u s in t e r p r e ta tio n c o u ld o c c u r . T h e r e f o r e , a m o r e r o b u s t m e a s u r e o f
311
m a x im u m c o n c e n t r a t io n i s r e q u ir e d , i . e . , a n a v e r a g e d v a lu e o f a ll s a m p lin g p o in t s a t e a c h tim e
in c r e m e n t, o f w h ic h th e h ig h e s t v a lu e is ta k e n . T h is v a lu e is d e fin e d a s th e m a x im u m m e a n f lo o r
c o n c e n tr a tio n , d e n o te d C f max. T h i s is b e lie v e d t o p r o v id e a b e tte r in d ic a tio n o f th e d e g r e e o f
s ta g n a tio n a n d w a s s u p p o r te d b y c lo s e r c o r r e la tio n s . In q u a n tif y in g th e m e a n f lo o r c o n c e n tr a tio n
( Cj ) e a c h m e a s u r e m e n t a t f lo o r le v e l w a s a v e r a g e d , w ith r e s p e c t t o a n a llo c a te d r e p r e s e n ta tiv e
flo o r a r e a (T a b le 4 3 ) , a s e q u a tio n ( 3 1 4 ) .
1 ( 6
(3 1 4 )
R m \ »=1
w here Ai is th e r e p r e s e n ta tiv e a r e a fo r a s p e c if ic s a m p lin g p o in t, a n d w a s c a lc u la te d a c c o r d in g to
th e r a d ia l d is t a n c e fr o m t h e r e le a s e , t h u s d if f e r in g v a lu e s d e p e n d in g u p o n w h e t h e r th e r e l e a s e w a s
a t t h e r o o m w a ll (b e n e a t h A H U in F ig u r e 1 8 6 ) o r t h e c e n tr e . O th e r a v e r a g in g t e c h n iq u e s w e r e
t e s t e d ( f o r e x a m p l e , s im p le n u m e r ic a l m e a n , e t c .) b u t w e r e f o u n d t o r e s u lt in m o r e s c a t t e r .
0.12 i
S ' 0.08
too
5 °
O 0/ 0,0
(J 0.04
0.00
0.00 0.04 0.08 0.12
f/jB (kgm -3)
Figure 103: Comparison o f maximum and mean maximum concentrations (V2 = 0.86)
A s s u m in g s y m m e tr y , w h e r e s a m p lin g p o in ts w e r e n o t lo c a te d (o n t h e r ig h t-h a n d s id e o f th e r o o m
w ith r e s p e c t t o th e o u tle t d ir e c tio n o f th e fa n ), v a lu e s fr o m th e le ft-h a n d s id e o f th e r o o m w e r e
u s e d a s a p p r o x im a tio n . A r e la tio n s h ip b e tw e e n Cfmax and Cf max w a s fo u n d (F ig u r e 1 0 3 ) , a n d
th e b e s t - f it w a s a c c o r d in g t o e q u a tio n ( 3 1 5 ) .
C f m a x ~ l - 2 8 - C ^ - max (3 1 5 )
312
3.1.3 Interpretation of Results
T h is s e c t io n p r e s e n t s t h e in t e r p r e ta tio n o f t h e s m o k e /v is u a lis a t io n t e s t s a n d th e r e s u lt s o f
c o n c e n tr a tio n m e a s u r e m e n ts . T h e c o m p le te s e t o f e x p e r im e n ta l c o n d itio n s a n d r e s u lts a r e
p r o v id e d in T a b le 5 8 ( A p p e n d ix 2 ) , a n d s e le c t e d d a t a is c o r r e la t e d t o c h a r a c t e r is e t h e in f lu e n c e o f
v a r i a b l e s o n r o o m c o n c e n t r a t i o n s . O b s e r v a t i o n s a n d t r e n d s a s s i s t e d in d e v e l o p i n g t h e Q R A a n d
th e a s s o c ia t e d e m p ir ic a l f o r m u la a n d f o r e s t a b lis h in g d e s ig n r e q u ir e m e n ts t o m in im is e f la m m a b le
c o n c e n tr a tio n s , a n d fo r th e fla m m a b le v o lu m e c a lc u la tio n s . T h e f o llo w in g e x a m in e s th e e f f e c t o f
im p o r t a n t v a r ia b le s in te r m s o f e f f e c t o n C f max . T o h e lp u n d e r s ta n d th e d is p e r s io n p r o c e s s ,
r e s u lt s a r e p r e s e n t e d in th r e e fo r m a t s : t im e - c o n c e n t r a t io n p r o f il e s f o r e a c h s a m p lin g p o in t , h e ig h t -
c o n c e n tr a tio n p r o file s a t p r e -s e t tim e in te r v a ls (w h e r e h e ig h t a n d d ir e c tio n o f u n it a ir flo w ,
r e f r ig e r a n t r e le a s e a n d m e a n c o n c e n t r a t io n a r e in d ic a te d ), a n d o v e r a ll tr e n d s in C fm ax. A ny
o b s e r v e d tr e n d s a r e c la r ifie d w ith b e s t - f it c u r v e s , a n d th e a s s o c ia t e d r e g r e s s io n c o e f f ic ie n t (r :) is
a ls o p r o v id e d t o in d ic a te th e q u a lit y o f t h e c o r r e la tio n . P r io r t o p r e s e n t in g th e r e s u lt s , q u a lita tiv e
in fo r m a tio n g a in e d fr o m v is u a lis a tio n t e s t s is d is c u s s e d .
entrainment into mixing from

return airflow discharge airflow
Figure 104: Observations from smoke-only tests
Visualisation test results

V is u a lis a t io n t e s t s w e r e c o n d u c t e d in t w o s ta g e s : in it ia lly w it h n e u t r a lly b u o y a n t s m o k e , a n d
s e c o n d ly w ith a m ix tu r e o f s m o k e a n d c a r b o n d io x id e t o in c r e a s e th e b u lk d e n s ity o f th e s m o k e .
T h e s m o k e o n ly t e s t s p r o v id e o b s e r v a t io n o f th e a ir flo w s g e n e r a t e d b y th e a ir h a n d le r , w h e r e a s
t h e s m o k e /C C T t e s t s in d ic a t e d t h e e f f e c t o f t h e u n it a ir flo w 7o n t h e d e n s e g a s r e le a s e .
313
entrainment into mixing from
return airflow discharge airflow COysmoke
Figure 105: Observations from smoke-C02 tests
T h e m a in o b s e r v a t io n s f r o m th e s m o k e - o n l y r e le a s e s w e r e th a t a r e le a s e w it h in t h e a ir s tr e a m o r
n e a r t h e i n l e t g r i l l e d r e w t h e s m o k e d i r e c t l y i n t o t h e r e t u r n a i r s t r e a m , c a u s i n g l o c a l i s e d m i x i n g in
t h e d is c h a r g e j e t . T h is th e n te n d e d t o m ix w it h t h e r o o m a ir w it h in th e f a r s id e o f t h e r o o m .
S im ila r ly w h e n th e r e le a s e w a s m a d e fr o m b e n e a th t h e A H U it w a s s e e n th a t a la r g e p r o p o r tio n o f
t h e s m o k e w a s r a p id ly d r a w n u p w a r d s t o t h e A H U a ir in le t g r ille , a g a in d is c h a r g in g a j e t o f
s m o k e , w h i c h t h e n d i l u t e s t o w a r d s f a r s i d e o f t h e r o o m . T h e s e o b s e r v a t i o n s a r e i l l u s t r a t e d in
F ig u r e 1 0 4 . A t a ll f ix e d f a n s p e e d s th e a ir h a n d le r m a n a g e d t o r a p id ly m o v e s m o k e t o v ir t u a lly a ll
p o in ts w ith in th e r o o m .
F o r s m o k e /C 0 2 r e le a s e s , s im ila r e v e n ts t o th e s m o k e o n ly t e s t s w e r e o b s e r v e d . O b s e r v a t io n s
g e n e r a lly s h o w e d th r e e d is t in c t ly d if fe r e n t m ix in g p r o c e s s e s o c c u r s im u lt a n e o u s ly : e n tr a in m e n t o f
a ir in t o t h e g a s p lu m e d u r in g it s d e c e n t , g a s e n t r a in m e n t in t o t h e r e t u r n a ir flo w ' t o t h e in l e t g r ille ,
a n d m ix in g c a u s e d b y d is c h a r g e a ir flo w n e a r f lo o r le v e l. A r e le a s e fr o m b e n e a th th e u n it a p p e a r e d
t o d raw ' th e d e n s e s m o k e u p to w a r d s th e in le t g r ille . T h e m a in d if fe r e n c e th o u g h w a s th a t th e
a m o u n t o f s m o k e /G O : e n tr a in e d d ir e c tly in t o t h e A H U in le t g r ille fr o m th e d if f u s e r w a s n o t a b ly
r e d u c e d . T h e r e w a s a ls o a m o r e p r o n o u n c e d ‘s p r e a d in g ’ e ffe c t o f th e s m o k e , th a t f a lls to w a r d s
th e flo o r a n d th e n a c r o s s to w a r d s th e e x tr e m itie s o f th e r o o m . T h e s e e f f e c t s a r e in f lu e n c e d b y th e
h ig h e r d e n s it y o f t h e p lu m e . A s t h e c lo u d flo w 's a c r o s s t h e f lo o r , t h e r e tu r n a ir e n t r a in s a
p r o p o r t io n o f t h e s m o k e in a c o u n t e r - f lo w m a n n e r . T h e d is c h a r g e a n d d is p e r s io n o f t h e e n tr a in e d
a ir in t h e A H U d is c h a r g e a i r f l o w i s a s w i t h t h e s m o k e o n ly c a s e . T h e s e a r e i llu s t r a t e d in F ig u r e
1 0 5 . R e l e a s e s f r o m b e s id e t h e u n it b e h a v e d in a s im ila r m a n n e r t o t h o s e f r o m b e n e a t h t h e u n it.
T h e s e v is u a lis a t io n t e s t s a r e u s e f u l in t w o r e s p e c t s . F ir s t ly , t h e y o f f e r a n in d ic a t io n o f t h e r e g io n s
w h e r e h ig h e r c o n c e n t r a t io n s a r e e x p e c t e d . T h e d e n s e s m o k e is s e e n t o ‘la n d ’ a t t w o lo c a tio n s :
314
d ir e c tly b e n e a th a n d in fr o n t o f th e d if fu s e r , a n d to w a r d s th e fa r w a ll w h e n th e a ir f lo w w a s
s u f f i c ie n t t o e n tr a in it in t o t h e A H U , a n d t h r o w it a c r o s s t h e r o o m . T h e p la c in g o f s a m p le p o in t s
a t t h e s e l o c a t io n s is t h e r e f o r e im p o r t a n t. S im ila r ly , f r o m t h e e n t r a in m e n t t h e o r y m e n t io n e d in th e
lit e r a tu r e r e v ie w , th e v e lo c i t y c l o s e t o t h e g a s /a ir in t e r f a c e is a k e y f a c t o r , w h ic h p r o v id e s a n
in d ic a tio n a s t o n e c e s s a r y p o s it io n in g o f a n e m o m e t e r s . T h e s e c o n d b e n e f it fr o m t h e s e
o b s e r v a t io n s is t h e c o n t r ib u t io n t o w a r d s in te r p r e ta tio n o f t h e c o n c e n t r a t io n m e a s u r e m e n t s .
0.005 0.010 0.015 0.020

Concentration (kg m -3)
Figure 106: Concentration-time profiles fo r Figure 107: Concentration-height profiles for

reference test (no. 10) reference test (no. 10)
Reference test
I n itia lly , t h e r e s u lt s f o r a r e f e r e n c e t e s t (n o . 1 0 ) a r e p r e s e n t e d t o a c t a s a b a s is f o r c o m p a r in g t h e
in f lu e n c e o f d if f e r e n t v a r ia b le s in o t h e r e x p e r i m e n t s . T h e r e f e r e n c e t e s t w a s r e p e a t e d o n t w o m o r e
o c c a s io n s ( n o s . 2 4 a n d 3 0 ) t o c h e c k c o n s is t e n c y o f t h e r e s u lt s , a n d a ll th r e e t e s t r e s u lt s w e r e
w ith in ± 1 5 % o f th e a v e r a g e . F ig u r e 1 0 6 s h o w s a c o n c e n tr a tio n - tim e p r o file fo r th e r e fe r e n c e te s t,
a n d t h e le g e n d r e f e r s t o s a m p lin g p o s it io n s in T a b le 5 7 ( A p p e n d ix 2 ) . G e n e r a l c h a r a c t e r is t ic s
w e r e o b s e r v e d fr o m th is d a ta : im m e d ia te ly f o llo w in g th e s ta r t o f th e r e le a s e , c o n c e n tr a tio n s a t th e
flo o r , c lo s e t o th e p lu m e , te n d t o r is e q u ic k ly , w it h a m o r e g r a d u a l r is e a t p o in ts fu r th e r a w a y .
C o n c e n tr a tio n s a t 1 m a n d 3 m te n d t o h a v e a s m o o t h e r a n d s lo w e r in c r e a s e . A k e y o b s e r v a t io n is
t h e e r r a tic c h a n g e s in f lo o r c o n c e n t r a t io n s , w h ic h is a p a r t ic u la r c h a r a c t e r is t ic o f r e l e a s e s w it h
f o r c e d a i r f l o w . F r o m t h e c e s s a t i o n o f t h e r e l e a s e a l l c o n c e n t r a t i o n s t e n d t o w 7a r d s t h e m e a n r o o m
c o n c e n tr a tio n a n d th e p r o file s b e c o m e m u c h s m o o th e r .
315
3
o 60 s
~©~ 120 s
2 2
E
300 s
o
33
1 1
0 0
0 0.001 0.002 0.003 0.004 0.005 0 0.005 0.01 0.015 0.02 0.025
Concertration (kg m-3) Concertntfbn (kg m-3)
Figure 108: Concentration-height pro files fo r Figure 109: Concentration-height pro files for
small release mass large release mass
F ig u r e 1 0 6 s h o w s th e d e v e lo p m e n t o f f lo o r , m id h e ig h t a n d c e ilin g c o n c e n t r a t io n a t o n e m in u te
in t e r v a l fr o m in it ia t io n o f th e r e le a s e u n til a r e la t iv e ly s t e a d y s t a t e e x i s t s . In t h is t e s t , it w a s s e e n
th a t f o ll o w i n g t h e in itia tio n o f th e r e le a s e g a s f a lls to w a r d s th e f lo o r , b u t a ls o a p r o p o r t io n is
d ir e c te d t o w a r d s c e ilin g le v e l ( 6 0 s ). A s th e r e le a s e c o n t in u e s a ll c o n c e n t r a t io n s in c r e a s e ( 1 8 0 s )
u n til t h e s u p p l y o f g a s c e a s e s a t 2 1 0 s a n d a m o r e h o m o g e n o u s r o o m c o n c e n t r a t io n d e v e lo p s ( 2 4 0
s t o 3 6 0 s ) . D u r in g t h is p e r io d , f lo o r c o n c e n t r a t io n r e d u c e s a s it i s m i x e d w it h a ir in t h e u p p e r
p a r ts o f th e r o o m a n d a p p r o a c h e s th e h o m o g e n o u s c o n c e n tr a tio n ,
0.03
2 0.02 -
O
0.01
0.00
0.0 0.5 1.0 1.5 2.0 2.5
Release mass (kg)
Figure 110: Effect o f released mass on Cfmax (r2 = 0.96)
Effect of released mass

T h e e f f e c t o f t h e r e le a s e d m a s s u p o n t h e c o n c e n t r a t io n h a s b e e n e v a lu a t e d , a s s h o w n in F ig u r e
1 0 8 , w h e r e a lo w e r m a s s , b a s e d o n th e s a m e r e le a s e tim e , r e d u c e s C fmax. T h i s tr e n d w a s
316
e x p la in e d b y e n tr a in m e n t t h e o r y . W h e n a r e l e a s e o c c u r s , t h e r a te o f e n tr a in m e n t is g o v e r n e d b y
its b u o y a n c y a n d m o m e n tu m . V o lu m e f lu x o f t h e r e le a s e h a s a m in im a l in f lu e n c e o n p lu m e
v e lo c it y a n d th e q u a n tit y o f e n tr a in e d a ir is p r o p o r t io n a lly le s s f o r a h ig h e r r e le a s e d m a s s . T h u s , a
h ig h e r m a s s f l o w r e s u lt s in a h ig h e r c o n c e n t r a t io n a t t h e f lo o r , g iv in g a h ig h e r d e n s it y , m e a n in g
th a t it r e q u ir e s m o r e w o r k t o p r o d u c e t h e s a m e a m o u n t o f m ix i n g . T h is e f f e c t w a s o b s e r v e d b y
c o m p a r in g F ig u r e 1 0 8 a n d F ig u r e 1 0 9 , w h e r e th e lo w r e le a s e m a s s is s e e n t o h a v e a n a lm o s t
h o m o g e n o u s c o n c e n t r a t io n w it h in t h e r o o m th r o u g h o u t t h e r e le a s e d u r a t io n e x c e p t n e a r t h e sta r t.
C o n v e r s e ly , w it h a la r g e r e le a s e m a s s , th e c o n c e n t r a t io n w it h in t h e u p p e r p a r t o f t h e r o o m is
s i g n i f i c a n t l y l e s s th a n t h a t a t t h e f lo o r , r e p r e s e n t in g t h e r e s id u a l q u a n tity 7 o f g a s in a d e n s e c lo u d .
A lin e a r r e la tio n s h ip b e t w e e n r e le a s e a m o u n t o f g a s a n d C f max is s e e n in F ig u r e 1 1 0 . T h is is
c o n tr a r y t o th e c a s e o f n o a ir f lo w ( e .g .. C le a v e r e t a l, 1 9 9 4 , K a t a o k a e t a l, 2 0 0 0 b ) w h e r e th e r e is
a n e x p o n e n t ia l i n c r e a s e in Cf max w it h m a s s . It f o ll o w s th a t w h e r e r o o m a ir flo w s u f f i c ie n t t o
p r o d u c e v e r y e f f e c t iv e m ix in g , th e c o n c e n tr a tio n C f max « CRm - a t le a s t u n til t h e e f f e c t o f th e
a ir f lo w 7c e a s e s t o p r o v id e s u f f i c ie n t m i x i n g t o la r g e r r e l e a s e d m a s s e s , w h e r e th e t r e n d is e x p e c t e d
t o e r r t o w a r d s t h e e x p o n e n t ia l tr e n d a s i f t h e r e w e r e n o a ir f lo w . F in a lly , it is lo g ic a l f o r th e
in t e r c e p t o f t h e lin e - f it lin e t o b e s e t a t z e r o s in c e n o c o n c e n t r a t io n s c a n d e v e lo p w it h o u t a r e le a s e .
0.06
0.04
0.02
0.00
0 200 400 600
Time(s)
Figure 111: Concentration-time profile fo r release with no airflow
E f f e c t o f r o o m a ir m o v e m e n t
A n u m b e r o f e x p e r im e n ts w e r e c o n d u c t e d t o h e lp u n d e r s ta n d th e d e p e n d e n c y o f m ix in g o n a ir flo w
r a te s . D iff e r e n t A H U fa n s p e e d s e t tin g s w e r e u s e d ( lo w , m e d iu m a n d h ig h ) a n d in a d d itio n , a
v o lta g e tr a n s fo r m e r w a s u s e d t o r e d u c e th e s u p p ly v o lta g e t o fu r th e r lo w e r th e f lo w r a te s.
317
n o t p r o v id e a d d it io n a l r e d u c t io n s in Cf max . A l s o n o t e w o r t h y is th a t w h e r e n o f o r c e d a ir f lo w w a s
u s e d , m e a n a i r s p e e d w a s n e v e r b e l o w a b o u t 0 . 0 3 m s ' 1. T h i s w a s b e l i e v e d t o b e d u e t o a
c o m b in a t io n o f r e s id u a l a ir m o v e m e n t w it h in t h e r o o m , a n d t h e g e n e r a t io n o f a ir m o v e m e n t f r o m
t h e d is p la c e m e n t o f a ir b y t h e d e s c e n d in g p lu m e o f C O 2. O b s e r v a t io n s o f th e s t r o n g d e p e n d e n c e
o f m i x i n g o n a ir f lo w r a te w e r e c o n s is t e n t w it h a ll t h e s t u d i e s in t h e lit e r a tu r e r e v i e w , a n d in
p a r t i c u l a r t h e s e e f f e c t s w e r e d e m o n s t r a t e d e x p e r i m e n t a l l y i n t e r m s o f in w a r d d i s c h a r g e a n d
e x tr a c t v e n tila tio n , b y L e a c h a n d B lo o m fie d ( 1 9 7 3 ) a n d F le tc h e r a n d J o h n so n ( 1 9 8 6 ) .
0.06 t
0.04 -
B
3
S- 0 .0 2 -
0.00
0.0 0.1 0.2 0.3 0.4 0.5
"a* ( m s ' 1)
Figure 114: Effect o f AHU airflow rate on Cfniax (r2 = 0.85)
W h e r e a i r c o n d i t i o n e r s h a v e m u l t i - d i r e c t i o n a l a i r f l o w 7, t h i s c a n i m p a c t o n t h e e f f e c t i v e n e s s o f t h e
m ix in g . F ig u r e 1 1 5 a n d F ig u r e 1 1 6 p r e s e n t r e s u lt s w h e r e t h e a ir d is c h a r g e d u c t w a s r o ta te d ± 30°
a b o u t th e h o r iz o n ta l (ta k e n a s 9 0 ° t o th e r e le a s e d ir e c tio n ). W ith a d o w n w a r d j e t , th e m ix in g is
im p r o v e d , w h e r e a s a n u p w a r d j e t r e s u lt s in h ig h e r f lo o r c o n c e n t r a t io n s th a n w ith a h o r iz o n t a l
d is c h a r g e . T h is is a ttr ib u te d t o th e d o w n w a r d a ir flo w 7 c r e a t in g a h ig h e r a ir s p e e d n e a r th e
s t r a t if y in g c lo u d a t f lo o r le v e l. T h is e f f e c t is s e e n in F ig u r e 1 1 5 , w h e r e t h e f lo o r c o n c e n t r a t io n h a s
m u c h lo w e r v a lu e s th a n w h e n th e d is c h a r g e is h o r iz o n ta l (F ig u r e 1 0 7 ). W o r th n o t in g , is th a t th is
s e t o f c o n d it io n s r e s u lt e d in a u n iq u e s it u a t io n w h e r e th e m a x im u m c o n c e n t r a t io n is e n c o u n t e r e d
a t 1 m a b o v e th e flo o r , r a th e r th a n a t th e flo o r . H o w e v e r , b e c a u s e th e c o n c e n tr a tio n s a t 1 m
e ls e w h e r e w ith in th e r o o m w e r e lo w e r , Cf max w a s s till a t flo o r le v e l.
T h e e f f e c t o f u p w a r d d i r e c t e d f l o w 7, a s s e e n i n F i g u r e 1 1 6 , r e s u l t s i n m u c h h i g h e r f l o o r
c o n c e n t r a t io n s th a n t h e d o w n w a r d o r h o r iz o n t a l c a s e . T h e u p w a r d a ir flo w 7 is s e e n t o p r o d u c e a
m o r e s tr a tifie d la y e r a n d th e f lo o r c o n c e n tr a tio n c h a n g e s d r a m a tic a lly w ith tim e (ju m p in g b y 1 %
w ith in s e v e r a l s e c o n d s ). E x a m in in g th e a ir s p e e d m e a s u r e m e n ts s h o w th a t a lth o u g h th e m e a n
319
r o o m v e lo c it y is s im ila r in a ll c a s e s , t h e lo c a l v e lo c it y n e a r th e f lo o r is m u c h h ig h e r in t h e c a s e o f
th e d o w n w a r d d ir e c te d d u c t . F ig u r e 1 1 7 s h o w s t h e d e p e n d e n c y o f Cf max o n th e d u c t d is c h a r g e
a n g le .
o 60s □ 60s
O ' 120 s -o - 120 s
240 s c 240 s 1 m
300 s
0.000 0.005 0.010 0.015 0.005 0.010 0.015 0.020

Comcentration (kg m-3) Concentration (kg m -3)
Figure 115: Concentration-height profiles fo r Figure 116: Concentration-height profiles for

downward air discharge (test no 14.) upward air discharge (test no 15.)
B y e x t r a p o la t in g t h e s e o b s e r v a t io n s , it is a n t ic ip a t e d th a t d is c h a r g e a ir o r ig in a t in g f r o m f lo o r
le v e l w ill m a k e Cf max a lm o s t in d e p e n d e n t o f d is c h a r g e a n g le , s in c e b o th d is c h a r g e a n d r e tu r n
a ir f lo w s a r e a lr e a d y p r e s e n t in th e r e g io n w h e r e s t r a t if ic a t io n m a y o c c u r .
0.020
upward
0.015 -
o
£ 0.010 o
lU ' 0.005 -
downward
0.000
0.0 0.5 1.0 1.5 20 2.5
Angle adjacent to release direction (rad)
Figure 117: Effect o f air discharge angle on Cfniax (r2 = 0.82)
N o r m a lly , o c c u p ie d s p a c e s te n d t o c o n ta in th e r m a l s o u r c e s s u c h a s a p p lia n c e s a n d p e o p le . In
o r d e r t o o b s e r v e th e in flu e n c e o f th e r m a l c o n v e c t io n c u r r e n ts o n a d e n s e g a s r e le a s e , a n
320
e x p e r im e n t w a s c o n d u c t e d w ith ite m s d e s ig n e d t o im ita te th e e f f e c t o f th e r m a l lo a d fr o m p e o p le
a n d c o m p u te r s . A ty p ic a l o f f ic e w o r k e r p o p u la tio n (4 p e r 4 5 n r ) w a s lo c a te d w ith in th e te s t
c h a m b e r a n d a r e le a s e m a d e fr o m th e c e n tr e o f th e r o o m . H e a t o u tp u t fr o m p e o p le (a t s e d e n ta r y
w o r k ) w a s m im ic k e d w it h th e r m a l m a n ik in s , s e t a t 1 0 0 W a n d e a c h P C w ith 2 0 0 W o u tp u t. E a c h
w a s lo c a te d a t a p p r o x im a te ly 2 .2 m s p a c in g fr o m e a c h o th er .
Rm
3
□ 60s
120 s
2
E
780 s
0
0 0.01 0.02 0.03 0.04 0.05
Concentration (kg m-3)
Figure 118: Concentration-height profiles resulting from thermal loads (test no. 43)
T h e r o o m c o n c e n tr a tio n s r e s u ltin g fr o m th e r e le a s e a r e illu s tr a te d in F ig u r e 1 1 8 . I n itia lly th e f lo o r
c o n c e n t r a t io n in c r e a s e d r a p id ly a s i f u n d e r q u ie s c e n t c o n d it io n s , b u t o n c e th e r e le a s e s t o p p e d
f lo o r c o n c e n tr a tio n b e g a n t o r e d u c e , w h ils t t h e c o n c e n tr a tio n a t 3 m in c r e a s e d t o w a r d s CRm ,

u n lik e a s it u a t io n w it h n o o r m in im a l f o r c e d a ir flo w 7 ( e .g .. F ig u r e 1 1 2 ) , w h e r e c e i lin g
c o n c e n tr a tio n r e m a in e d n e a r z e r o . T h is d if f e r e n c e is d u e t o th e m ix t u r e a t lo w le v e l b e in g
e n tr a in e d u p w a r d s b y th e r m a l c o n v e c t io n . M e a n r o o m te m p e r a t u r e w a s 2 1 ,7 ° C w it h m a x im u m
v a r ia tio n s o f ± 1 .6 K - a r a n g e s im ila r t o o th e r te s ts w ith o u t th e r m a l s o u r c e s ( e x c e p t f o r fa n
m o t o r s ) . E v id e n t ly , o n ly m in o r d if f e r e n c e s in lo c a l a ir te m p e r a tu r e a r e n e e d e d t o g e n e r a t e
s u ffic ie n t a ir m o v e m e n t t o d is p e r s e a la r g e r e fr ig e r a n t r e le a s e .
Effect of post-ventilation period

T h e r e fe r e n c e t e s t u s e d a p o s t-v e n t ila tio n p e r io d th a t b e g a n a t th e s a m e tim e a s th e r e le a s e , a n d
fin is h e d a t th e s a m e tim e a s th e c e s s a tio n o f th e r e le a s e , i.e ., s im u lt a n e o u s ly fo r 3 .5 m in u te s . T h e
r e l a t i o n s h i p b e t w e e n m i x i n g a n d t h e p o s t v e n t i l a t i o n p e r i o d w ?a s e v a l u a t e d b y s t o p p i n g t h e f a n
p r e m a tu r e ly . F ig u r e 1 1 9 s h o w s th a t a s h o r te r p o s t - v e n t ila t io n p e r io d ( r e la tiv e t o r e le a s e t im e ) h a s
321
a d is p r o p o r t io n a t e e f f e c t in in c r e a s in g CJ max . I n fa c t b y s to p p in g th e a ir flo w h a lfw a y th r o u g h th e
r e le a s e p r o d u c e d v ir tu a lly n o d if fe r e n c e in C f max, th a n i f n o f o r c e d a ir f lo w w a s p r e s e n t a t a ll.
0.05 -r 0.04
0.04 1
<+■. 0.03 -
£
Pp 0.03 -
0.02
a
s 0.02 -
lo kj
o/
0.01 z
0.01 -
0.00 - 0.00
1 2 3 5 10 15 20
Post-verlflaton period (mins) Release tine (mins)
Figure 119: Effect ofpost-ventilation period Figure 120: Effect o f release time with 3.5
on Cfmax with 3.5 minutes release time. minute post-ventilation on Cf max
R e fr ig e r a n t le a k s w ill n o t n o r m a lly r e s u lt in a n in s ta n ta n e o u s r e le a s e , s o t h e le a k m a y p e r s is t f o r
lo n g e r t h a n 3 . 5 m in u t e s o f a c a t a s t r o p h ic r e l e a s e . T h e r e s u lt s p r e s e n t e d in F i g u r e 1 2 0 s h o w t h a t
w h e n t h e r e le a s e o v e r s h o o t s t h e p o s t - v e n t ila t io n p e r io d , s ig n if ic a n t ly h ig h e r C fmax d e v e l o p s . T h is
i s c o n s is t e n t w it h th e c a s e o f r e d u c in g p o s t - v e n t ila t io n tim e . In b o th s it u a t io n s th e r e is a la c k o f
a ir m o v e m e n t , t h e r e b y in h ib it in g m ix in g . A l s o n o t e w o r t h y is t h e e f f e c t o f d o u b lin g th e r e le a s e
d u r a t io n h a s v e r y lit tle im p a c t o n in c r e a s in g th e f lo o r c o n c e n tr a tio n . T h is d if f e r e n c e is u n d e r s t o o d
t o b e d u e t o t h e c o m p e n s a t o r y e f f e c t o f t h e lo n g e r t im e a v a ila b le f o r e n tr a in m e n t o f a s p e c if ic
r e le a s e d m a s s . E x a m in a tio n o f th e h e ig h t -c o n c e n tr a tio n p r o file s h e lp s t o fu r th e r c la r if y th e
s itu a t io n .
I n F ig u r e 1 2 1 , it is s e e n th a t in it ia lly d u r in g th e p o s t - v e n t ila t io n p e r io d th e C 0 2 is w e ll- m i x e d
t h r o u g h o u t t h e h e ig h t o f t h e r o o m . A f t e r c e s s a t i o n o f t h e a ir d is c h a r g e a t 2 1 0 s , t h e c o n t in u in g
r e l e a s e c a u s e s a r a p i d i n c r e a s e in c o n c e n t r a t i o n m a i n l y a t f l o o r l e v e l t h a t a p p e a r s t o b e n e g l i g i b l e
v a r ia tio n fr o m th e e n d o f th e r e le a s e a t 4 2 0 s to 1 1 4 0 s . T h e s e r e s u lts s u g g e s t th a t a lo n g e r p o s t
v e n t ila t io n p e r io d t o p r o v id e c o n tin u e d f o r c e d a ir s p e e d m a y b e n e c e s s a r y w h e r e s it u a t io n s o f
e x te n d e d r e le a s e tim e s w o u ld o th e r w is e d e v e lo p u n a c c e p ta b ly h ig h c o n c e n tr a tio n s .
322
3
□ 60s a 60s
~O ~ 180s
2 -A —300 s .^ I ^
-© -4 2 0 s 4
1140s
1 840s
1140s
0.005 0.01 0.015 0.02 0.025 0 0.01 0.02 0.03
Concentration (kg m-3) Concertration (kgm-3)
Figure 121: Concentration-height profile fo r Figure 122: Concentration-height profile for

a 7 minute release (test no. 29) a 14 minute release (test no. 25)
A r e le a s e o f 1 4 m in u te s f o llo w e d th e s a m e p a tte r n (F ig u r e 1 2 2 ) a lth o u g h th is w a s s lo w e r
p a r tic u la r ly w it h in th e la s t f iv e m in u te s w h e r e th e c o n c e n tr a tio n h a r d ly c h a n g e d . T h is s u g g e s t s
th a t a ft e r a c e r ta in tim e a p r o lo n g e d r e le a s e r e a c h e s a m a x im u m f lo o r c o n c e n tr a tio n a n d th e r e fo r e
c lo u d h e ig h t in c r e a s e s . A n o th e r o b s e r v a t io n is th a t th e m id - a n d c e ilin g le v e l c o n c e n tr a tio n s a r e
a p p r o x im a t e ly t w i c e a s h ig h f o r t h e 7 m in u t e r e l e a s e c o m p a r e d t o t h e 1 4 m in u t e r e l e a s e . T h is is
a tt r ib u te d t o t h e h ig h e r m a s s f l o w ( f o r t h e 7 m in u t e r e le a s e ) b e in g a s w e ll m ix e d a s t h e lo n g e r
r e le a s e , t h u s d e v e lo p in g a n in itia l c o n c e n t r a t io n t w i c e a s h ig h .
------ q.----- 9 ------- a ------ --------------------------- .

a 60s \ 1 □ 60s
-©• 120 s ; M -o ~ 1 2 0 s |
180 s Fa It - * - 1 8 0 s
-©- 240 s -© -240 s
I I If; -* -3 0 0 s
-* -3 0 0 s \ ] jji - i —360 s
—♦—360 s
□
\ \ \
yhr
---------------- r-E3----------- \ v----- ----------------------+1
0.00 0.02 0.04 0.06 0.08 0.000 0.005 0.010 0.015 0.020
Concentration (kgm-3) Concentration (kg m-3)

low release height (test no. 20) low release and discharge height (test no. 21)
323
E f f e c t o f r e le a s e a n d a ir d is c h a r g e p o s it io n
T h e e f f e c t o f d iffe r e n t r e le a s e h e ig h ts is k n o w n t o b e a s ig n ific a n t fa c to r w h e n th e r e is n o fo r c e d
c o n v e c t i o n , a s o b s e r v e d f r o m t h e l i t e r a t u r e r e v i e w 7. T e s t s w e r e c o n d u c t e d t o h e l p d e t e r m i n e t h i s
e f f e c t w h e n th e r e is fo r c e d a ir flo w . F ig u r e 1 2 3 s h o w s th e c o n c e n tr a tio n -h e ig h t p r o file s fo r a
r e l e a s e h e i g h t o f 0 . 4 m . C o m p a r i n g t h i s t o t h e r e f e r e n c e t e s t ( F i g u r e 1 0 7 ) it i s s e e n t h a t
s ig n ific a n tly g r e a te r c o n c e n tr a tio n s o c c u r a t th e flo o r . A ls o , c o n c e n tr a tio n s a t th e u p p e r p a r t o f
th e r o o m , 0 .0 0 4 k g m ° , a r e h a lf th a t o f th e r e fe r e n c e te s t. A tr e n d s h o w in g th e d e p e n d e n c e o f
C f,max on r e le a s e h e ig h t is s h o w n in F ig u r e 1 2 5 th a t s u g g e s t s a n e x p o n e n t ia l in c r e a s e a s th e
r e le a s e p o s itio n a p p r o a c h e s th e flo o r . T h is tr e n d is e x p la in e d b y th e d e s c e n d in g p lu m e h a v in g le s s
e n tr a in m e n t w it h t h e s u r r o u n d in g s i f r e le a s e d c lo s e r t o t h e f lo o r , a n d o n c e a t f lo o r le v e l t h e
c o n c e n tr a tio n a n d th e r e fo r e th e m ix t u r e d e n s it y is h ig h e r th a n i f it h a d b e e n r e le a s e d c lo s e r t o th e
c e i lin g , m a k in g it l e s s e a s y t o d is p e r s e . G iv e n t h a t h ig h r o o m a ir s p e e d h a s a m a j o r in f lu e n c e o n
t h e p l u m e ' s r a t e o f e n t r a i n m e n t it i s b e l i e v e d t h a t t h e d e p e n d e n c e o f C f max o n r e le a s e h e ig h t is
s tr o n g e r th a n u n d e r q u ie s c e n t c o n d itio n s .
0.08 0.08 T
0.06 0.06
£ £
ob pb
^ 0.04 0.04
lU ' 0.02 0.02
0.00 0.00
0.0 0.5 1.0 1.5 2.0 2.5 0.0 0.2 0.4 0.6 0.8 1.0
Release height (m) Ratio o f rebase to air discharge height (-)
Figure 125: Effect o f release height on Cfmax Figure 126: Effect o f release height and unit
(r = 0.96) height (h /h ^ on Cfmax ( r = 0.93)
T h e e f f e c t o f r e le a s e h e ig h t w a s in v e s t ig a t e d fu r th e r b y lo w e r in g a ir d is c h a r g e h e ig h t , t h e r e b y
m o v in g th e h ig h d is c h a r g e v e lo c ity 7c lo s e r t o th e s p r e a d in g p lu m e . C o m p a r in g F ig u r e 1 2 3 a n d
F ig u r e 1 2 4 w h e r e a ir d is c h a r g e is n e a r t h e r e l e a s e h e ig h t , c o n c e n t r a t io n p r o f il e s a r e s e e n t o
d e v e lo p d if f e r e n t ly . W h e n t h e a ir d is c h a r g e is c l o s e t o t h e f lo o r th e r e i s a r e la t iv e ly s l o w e r r is e in
f l o o r c o n c e n t r a t i o n a n d it i s c l o s e l y f o l l o w e d b y t h e c o n c e n t r a t i o n s i n t h e u p p e r p a r t s o f t h e r o o m ,
e v e n tu a lly m o v in g to w a r d s CRm b y th e e n d o f th e r e le a s e . T h is e v e n t is e x p la in e d b y th e lo w -
324
le v e l a ir s tr e a m r a p id ly e n tr a in in g th e g a s c l o s e t o t h e f lo o r a n d t h e n c o n t in u in g u p th e f a r w a ll
to w a r d s th e c e ilin g .
F ig u r e 1 2 6 s h o w s a d d itio n a l r e s u lt s th a t a c c o u n t fo r t h e u n it a t d if f e r in g h e ig h t s , w h e r e d a t a is
d is p la y e d fo r v a r io u s r a tio s b y v a r y in g b o th th e r e le a s e h e ig h t t o a ir d is c h a r g e h e ig h t, (h r lh d ).
O th e r th a n th e e f f e c t o f th e a c tu a l r e le a s e h e ig h t, th is a ls o in d ic a te s im p r o v e d m ix in g o f th e
r e f r ig e r a n t a s t h e r e le a s e p o s it io n a p p r o a c h e s t h e a ir d is c h a r g e . T h e s e o b s e r v a t io n s a r e e x p la in e d
b y a c o m b in a t io n o f h ig h e r a ir s p e e d p r o m o t in g e n tr a in m e n t w it h t h e s p r e a d in g p lu m e . W h e r e
b o t h r e le a s e a n d a ir d is c h a r g e is a t a m e d iu m h e ig h t (hr/hd ~ 0 . 8 ) a c o m p a r a b le r e s u lt t o th e
c a s e s a t h ig h a n d lo w h e ig h t s w a s f o u n d . T h is im p lie s th a t t h e h ig h lo c a l a ir s p e e d s n e a r t h e a ir
d is c h a r g e a n d r e tu r n d u c t s c r e a t e s s ig n if ic a n t a m o u n t s o f m ix i n g w h e n in c l o s e p r o x im ity ' t o t h e
r e le a s e s o u r c e . A s w ith th e e a r lie r d is c u s s io n , th e d e p e n d e n c e o f d is p e r s io n u p o n a ir s p e e d r a th e r
t h a n v o lu m e t r ic f l o w r a te is h ig h lig h t e d . It a l s o f o l l o w s th a t i f a ir s p e e d is lo w e r e d , th e b e n e f it o f
h a v in g a n a ir d is c h a r g e p o s it io n c lo s e t o th e r e le a s e s o u r c e le s s e n s . S im ila r ly , w it h a ir s p e e d
in c r e a s e d , its e f f e c t o n m ix in g w o u ld b e s u ffic ie n t fo r C f max —» CRm, a g a in n e g a tin g th e e ffe c t o f
th e p r o x im it y o f a ir d is c h a r g e t o r e le a s e s o u r c e .
0.03
0.02 -
00
! o.oi
0.00
0 4 6 8
Horizontal distance (m)
Figure 127: Effect o f release location (horizontal distance from air disch.) on Cfytnax ( r = 0.93)
A n u m b e r o f t e s t s w e r e c o n d u c t e d t o d e te r m in e t h e e f f e c t o f t h e r e le a s e o c c u r r in g a t d if fe r e n t
lo c a t io n s w it h in t h e r o o m . T h e a lte r n a te lo c a t io n s w e r e o n t h e w a l l b e s id e t h e e n d o f t h e u n it a n d
in t h e o p p o s i t e c o m e r f r o m t h e a ir d is c h a r g e t o r e p r e s e n t a l o c a t i o n a s f a r a d is t a n c e f r o m t h e
o r ig in a l le a k p o s itio n a s p o s s ib le . F ig u r e 1 2 7 s h o w s c h a n g e in C f max w ith r e s p e c t to h o r iz o n ta l
d is ta n c e o f th e r e le a s e fr o m th e c e n tr e o f th e A H U . It a p p e a r s th a t th e lo c a tio n h a s a s ig n ific a n t
e f f e c t o n th e c o n c e n tr a tio n w h e n th e r e le a s e p o in t w a s lo c a te d a t th e o p p o s it e c o m e r o f th e r o o m .
325
T h is is d e e m e d t o b e a r e s u lt o f lo w e r lo c a l a ir s p e e d n e a r th e d e s c e n d in g th e p lu m e - a n a lo g o u s
t o t h e c a s e o f l o w e r i n g r e l e a s e h e i g h t b e n e a t h t h e a i r d i s c h a r g e . S i m i l a r l y , in t h e a b s e n c e o f
a ir flo w C f max w o u ld b e c o m e in d e p e n d e n t o f t h e ( h o r iz o n t a l) r e le a s e lo c a t io n , a n d i f h ig h a ir
s p e e d s w e r e p r e se n t th r o u g h o u t th e r o o m , Cf max —> CRm r e g a r d le s s o f r e le a s e lo c a tio n . O n th is
b a s is it m a y s u g g e s t th a t f o r a n y g i v e n s e t o f c o n d it io n s , th e r e c o u ld b e a m e a n a ir s p e e d th a t
w o u l d r e s u l t i n a ‘‘p e a k ” s e n s i t i v i t y o f a i r d i s c h a r g e a n d r e l e a s e p o s i t i o n o n C J max .
Figure 128: Configuration o f AHU airflows showing discharge and return air paths
E q u ip m e n t ty p e
T w o a d d itio n a l A H U s a n d t w o r e fr ig e r a te d d is p la y c a b in e ts w e r e a ls o te s te d . T h e r e s u lt s fr o m
t h e s e t e s t s c a n n o t b e d ir e c tly c o r r e la te d a lo n g s id e t h o s e f o r A H U # 1 b e c a u s e th e ir c o n f ig u r a t io n s
d if fe r e d , p a r tic u la r ly in te r m s o f a ir f lo w c h a r a c t e r is t ic s a n d p o s it io n in g . P r im a r ily , t h e p u r p o s e o f
t e s t in g d if fe r e n t e q u ip m e n t w a s t o e s t a b lis h w h e th e r t h e tr e n d s o b s e r v e d w it h A H U # 1 c o u ld b e
a p p lie d a c r o s s a r a n g e o f e q u ip m e n t s o th a t t h e f in d in g s m a y b e c o n s id e r e d g e n e r ic . A s c h e m a t ic
d ia g r a m o f th e a ir f lo w c h a r a c t e r is t ic s o f t h e th r e e A H U s a n d t h e r e f r ig e r a te d c a b in e t s a r e
p r o v id e d in F ig u r e 1 2 8 a n d F ig u r e 1 2 9 r e s p e c t iv e ly . T h e a r r o w p o in t in g a w a y fr o m t h e u n it
r e p r e s e n ts th e d is c h a r g e a ir a n d t h o s e p o in t in g in t o th e u n it , t h e r e tu r n , o r in le t a ir f lo w s .
T h e p h y s ic a l c o n s t r u c t io n o f A H U # 2 d if f e r e d s lig h t l y f r o m A H U # 1 in th a t t h e d ir e c t io n o f a ir
d is c h a r g e a p p e a r s t o b e s p a n n e d a c r o s s s o m e 6 0 ° r a th e r th a n d is c h a r g in g d ir e c t ly in o n e
d ir e c tio n . A n o th e r d if f e r e n c e b e tw e e n th e m w a s th a t its in le t g r ille w a s p o s it io n e d b e l o w t h e
d is c h a r g e d u c t, w h e r e a s th e fo r m e r w a s o p p o s it e w ith th e d is c h a r g e a t th e to p . A H U # 3 d iffe r e d
in a n u m b e r o f w a y s ; o t h e r t h a n b e i n g l o c a t e d a t t h e c e i l i n g r a t h e r t h a n t h e w a l l , i t h a s f o u r
d o w n w a r d - p o in t in g d is c h a r g e g r ille s a r o u n d t h e p e r im e te r o f t h e u n it, w it h t h e in le t in t h e c e n tr e .
326
display area
Cabinet # 1 Cabinet #2
Figure 129: Configuration o f cabinet air flows
I n c o m p a r in g A H U # 1 p e r f o r m a n c e in F i g u r e 1 0 7 a g a in s t A H U # 2 in F ig u r e 1 3 0 w h e r e t h e
c o n d it io n s w e r e id e n t ic a l t o th e r e f e r e n c e t e s t , A H U # 2 p r o d u c e d b e tt e r m ix in g th a n in th e
r e fe r e n c e c a s e . D e s p it e A H U # 2 h a v in g a lo w e r v o lu m e f lo w r a te th a n A H U # 1 , th e c o n c e n tr a tio n
a t c e s s a tio n o f th e r e le a se w a s c lo s e r to CRm, and C fmax w a s lo w e r . T h is d if f e r e n c e is c o n s id e r e d
t o b e a r e s u lt o f t w o fa c to r s . F ir s t, th e d is c h a r g e d u c t a r e a o f A H U # 2 w a s p r o p o r tio n a lly le s s
(fo r th e lo w e r v o lu m e f lo w r a te ), w h ic h p r o d u c e d a h ig h e r m e a n a ir s p e e d w ith in th e r o o m . T h e
s e c o n d is th a t th e g e o m e tr y o f d is c h a r g e d u c t i t s e l f p a r tia lly s p a n n e d d o w n w a r d s , r a n g in g fr o m
a b o u t 0 ° t o a b o u t 3 0 ° b e lo w th e h o r iz o n ta l. A ir s p e e d m e a s u r e m e n ts s u p p o r te d b o th o f th e s e
o b s e r v a t io n s , s h o w in g a s ig n if i c a n t ly h ig h e r a ir s p e e d c l o s e r t o t h e f lo o r , c o n s is t e n t w it h t h e t e s t
o f A H U # 1 w it h d o w n w a r d d ir e c te d d is c h a r g e .
a 60s a 60s
■*- 120 s - o 120 s
240 s
0.000 0.005 0.010 0.015 0.020 0.00 0.01 0.02 0.03

Concentration (kgm-3) Concentration (kgm -3)

AHU #2 (test no. 32) AHU #3 (test no. 36)
327
A H U # 3 a l s o d e m o n s t r a t e d g o o d m i x i n g c h a r a c t e r i s t i c s . I n F i g u r e 1 3 1 it i s s e e n t h a t t h e h i g h e s t
c o n c e n tr a tio n d e v e lo p s a b o v e th e f lo o r a n d e v e n tu a lly a p p r o a c h e s CRm . T h e h ig h a ir s p e e d
m e a s u r e d c lo s e t o t h e f lo o r , p a r t ia lly a r e s u lt o f t h e d o w n w a r d a ir d is c h a r g e , is r e s p o n s ib le fo r t h e
r a p id m ix in g . A p e r ip h e r a l a ir d is c h a r g e a n d c e n t r a l r e tu r n d u c t is b e lie v e d t o c a u s e t h e h ig h
c o n c e n tr a tio n s a t 1 m , w h e r e th e e n tr a in e d g a s is d r a w n b a c k u p a c r o s s t h e s a m p lin g p o in t s
w ith in t h e m id d le o f th e r o o m . I n te r e s tin g ly , th is is th e o n ly o th e r s it u a t io n w h e r e t h e m a x im u m
lo c a l c o n c e n tr a tio n is a t 1 m in s te a d o f a t flo o r le v e l, in a d d itio n t o A H U # 1 w h e n d is c h a r g e
a ir flo w r is in d o w n w a r d d ir e c tio n .
B o th r e fr ig e r a te d c a b in e ts w e r e flo o r -m o u n te d (F ig u r e 1 2 9 ). C a b in e t # 1 is a n o p e n fr o n te d
d is p la y c a b in e t w it h t h e c o n d e n s in g u n it lo c a t e d b e n e a th th e s t o r a g e s p a c e , w it h t h e a ir flo w 7b e in g
p r o d u c e d b y th e c o n d e n s e r f e n , d r a w in g a ir th r o u g h f r o m th e r e a r o f t h e u n it a n d d is c h a r g in g a t
t h e fr o n t . T h e C a b in e t # 2 h a s a t o p - m o u n t e d c o n d e n s in g u n it t h a t is o n ly o p e n t o t h e t o p , a g a in
w it h a f e n d r a w in g a ir d o w n th r o u g h t h e c o n d e n s e r a n d th e n d is c h a r g in g it o u t in a n u p w a r d s
d ir e c tio n . In c o m p a r is o n t o t h e A H U s , t h e m o s t im p o r t a n t d if f e r e n c e is t h e lo w a ir v e lo c it y a t
d is c h a r g e , d u e t o a r e la t iv e ly la r g e d u c t a r e a .
3
D 60s
- o 120 s
-* r - 180 s
2 -©-240 s
-* -3 0 0 s
- + -4 8 0 s
1
0 0 4&
0.00 0.02 0.04 0.06 0.08 0 0.005 0.01 0.015 0.02
Concentration (kgm -3) Concentration (kg m-3)
Figure 132: Concentration-height pro files fo r Figure 133: Concentration-height profiles fo r

cabinet #1 (test no. 40) cabinet #2 (test no. 41)
T h e f l o o r c o n c e n t r a t i o n s r e s u l t i n g f r o m a r e l e a s e b e s i d e t h e C a b i n e t # 1 w 7c r e r e l a t i v e l y h i g h , a s
s h o w n in th e p r o f ile s in F ig u r e 1 3 2 . T h is is b e c a u s e a ir d is c h a r g e v e lo c it y - a n d t h e r e fo r e m e a n
r o o m a ir s p e e d - is lo w , a n d th e r e le a s e s o u r c e b e in g a t lo w le v e l. H o w e v e r , w h ils t a h ig h
c o n c e n tr a tio n d e v e lo p e d a t th e b e g in n in g o f th e r e le a s e , th e c o n c e n tr a tio n a t flo o r le v e l h a lv e s b y
t h e e n d o f th e r e le a s e a n d p o s t - v e n t ila t io n p e r io d . T h is s u g g e s t s th a t w h ils t t h e a ir v e lo c it y 7 is lo w ,
328
the effect o f air movement still contributes to mixing, similar to the case where thermal sources
were used.
Table 44: Information on test parameters
Parameter Observ ation for lower C f max Test range
Refrigerant release mass Lower mass 0.3 - 2.0 kg

High air speed, counter-flow to From integral fan only, uRm= 0
Room air movement
release plume - 0.45 m s'1
Air discharge height Low. close to release 0.2-2.5 m
Horizontal. 7u!3 rad, 2• tt/3
Air discharge direction Downward
rad
Post-ventilation time Longer duration 0 - continuous
High, close to air discharge and
Refrigerant release height 0.2 -2 .5 m
return
Close to air discharge and Beneath fan, other points in
Refrigerant release location
return room
Refrigerant release time Longer duration 210-840 s
Room size Smaller per unit mass releaseda 36 - 135 m3
Room tightness High infiltration1 Fully sealed
Thermal internal convection Increased thermal source output None
Room air temperature Negligible 1 Fixed (20°C - 22°C)
Room obstacles Can reduce C f 1 Not tested
Release phase Liquid T Vapour

Refrigerant release velocity High * 0.03-0.17 m s'1
Refrigerant release direction UpwardT Downward
t Based on literature review.
As with Cabinet #1, Cabinet #2 also produced relatively high floor level concentrations probably
due to low air speeds within the room, Figure 133. Following cessation of the release, the floor
level concentration begins to reduce slightly, whilst the mid and ceiling level concentrations
continue to increase. With AHU #1, when hr / hd —» 1, C f max was mainly independent o f release
height, whereas Cabinet # 1 produced a much higher concentration than Cabinet #2, despite both
their hr I hd ~ 0.8. If the airspeed associated with both cabinets is compared to that of AHU #1, it
is found to be much lower, approaching that of a still room. Therefore the influence of air
discharge height becomes less relevant, and this results in a stronger dependency of C f max on
release height.
329
Concluding remarks
Certain variables have been identified that characterise the development of floor concentration
due to a gas release from RAC equipment. It was found that C fmax always occurred soon after
cessation of the release (for example. Figure 106 and Figure 111). Once C fmax was reached, ex
tended to reduce gradually towards a homogenous mixture. The conditions leading to improved
mixing and lower floor concentrations are:
i smaller refrigerant charge, bearing in mind that it should be limited so that it does not exceed
the LFL when evenly mixed within the room,
ii high mean room air speed, achieved by increasing fan airflow rate or reducing outlet area,
iii directing air discharge downwards,
iv having a release source at high level, by positioning refrigerant containing parts as far above
the floor as possible,
v installing points of potential release close to air discharge/inlet,
vi using a longer post-ventilation period.
A summary of the key test parameters and the corresponding conditions favourable for reducing
C f^ are listed in the first and second columns o f Table 44, and additional parameters not
tested, but where detailed assessments were found in the literature review are also included. The
third column lists the range of each variable tested. Despite using these techniques as an aid to
reduce C f max, local concentrations above the LFL will always be present in the event of a leak.
However, by incorporating appropriate measures based on the observed trends, the flammable
region can evidently be minimised. This observation applies to different equipment when the
dispersion characteristics and conditions were comparable or similar.
3.2 Development of correlation for floor level concentration

It was necessary7to develop a correlation so that the concentration following a release could be
predicted quantitatively. The correlation should be applicable to as broad range of RAC
equipment as possible. An advantage of a generic formula is that it can be used to determine
numeric design limits for RAC equipment to minimise the occurrence of concentrations exceeding
the LFL following a refrigerant release.
330
The floor concentration following a release is a function of various characteristics associated with
the design and installation of that equipment. Some of these w ere identified as independent
variables, and some as dependent variables. The independent variables are those that can be fixed
within the design and construction of the equipment. Dependent variables are those that cannot be
controlled by equipment design. For example, refrigerant release velocity is dependent, whereas
release height is an example of an independent variable. Although room size is a dependent
variable it is assumed to be controlled through correct installation practice, and therefore is
handled as an independent variable. Where a dependent variable is considered, its value w'as fixed
according to what would produce highest C f max . A correlation was developed for two separate
cases: firstly, determination of C f mjx resulting from a catastrophic release, and secondly,
determination of C f development and C f max for a prolonged release.
3.2.1 Maximum floor concentration from catastrophic release

Within the literature, several methods have been employed to estimate concentrations from the
release of gases. These include: computational fluid dynamics (CFD), zonal decay models
(ZDM), Gaussian (concentration field probability) models, non-dimensional correlations, and
purely empirical correlations. It is considered that simplicity’, ease of use, and generic application
are prerequisites for the chosen approach. This directs towards a non-dimensional correlation
being most suitable, and this approach has also been successfully employed for a similar purpose
elsewhere (i.e., Cleaver at al, 1994).
Dimensional analysis 25
The procedure to develop the correlation initially required identification of the variables that
significantly affect C f max . Each w’as assigned a quantifiable variable that could be employed for
the correlation. From the list in Table 44, the key dependent variables are listed in Table 45.
Some of the variables in Table 45 were combined to reduce the number of non-dimensional
groupings. A reduced gravity’ ( g ' ) is defined to represent the buoyancy of the refrigerant relative
to the air, equation (316).
g '=g . ^ z £ ± (316)
Pa
"5 Note that the correlations presented here differ slightly from those in Colboume and Suen (2003b) due
to improvements in data analysis and availability of new’validation data.
331
where g is gravitational acceleration, p r is density o f the released gas, and p a is the density o f
air (based on room temperature and atmospheric pressure).
Table 45: Summary o f influencing variables
Parameter Selected variable Units

Concentration Max mean floor concentration kgm '3
Quantity of released refrigerant Mass kg
Choice of refrigerant Density kgm '3
Room air movement Volume flow rate m3 s-1
Air discharge height Height m
Air discharge direction Duct angle rad
Air discharge area Duct area nr
Refrigerant release height Height m
Room size Area. Height m2. m
Room air Density kg m'3
Since entrainment of a release into the air is normally a function of velocity, air volume flow rate
from the unit and duct discharge area were combined to obtain the mean room air speed ( uRm).
Other studies in the literature demonstrate the relevance of uRm, for example, the experiments by
Leach and Bloomfield (1974), where the air speed was adjusted by altering air inlet vent area,
whilst maintaining volume flow rate resulted in different concentrations. The equation for
calculating uRm given by Etheridge and Sandberg (1996) was compared against averaged air
speeds from the experiments. Figure 134. In averaging the local air speeds, different weightings
were applied to each sensor and the best correlation was obtained with based on the relative
volumes listed in Table 57 (Appendix 2). The sensors at 2 m and 3 m above floor level were
neglected because the small number of sensors at these positions was considered insufficient to
fully account for the variations in air speed at those horizontal planes. More so, it was deemed
acceptable to neglect these since most mixing occurs close to floor level. The comparison showed
measured air velocities correlated well with the calculated uRmusing equation (317) o f Etheridge
and Sandberg, where the discharge coefficient, cd =0.45.

where, Vd is volume flow rate of air discharged from the duct, A d is outlet area of duct
discharge and VRm is room volume.
0.5 t
I 0.4
Is 03 o/v'O
o
•o
a
1& 02 H
0>
2 O0
0.1
0.0
0.0 0.1 0.2 0.3 0.4 0.5
C alculated (m s " 1)
Figure 134:Comparison o f measured room air speed (r2 = 0.92)
A number o f non-dimensional groupings were established according to the Buckingham pi-

theorem, for the independent variables identified in Table 45. Initially, only measurements
obtained for AHU #1 were used. There are six variables, which comprise three fundamental
dimensions (mass, length and time) meaning that three repeating variables are required for the
dimensional analysis, leaving three pi-groups to be evaluated. The pi-groups are formed as
follows (equations 318, 319 and 320).
n , = ---------------- (318)
iARm'hRm) '
FL = g ’’K (319)
n3 (320)
First, n , represents the ratio of maximum floor concentration to mean room concentration,
termed mixing effectiveness ( s ); as s nears unity, homogenous mixing is observed. Secondly,

n 2 follows the definition of the Richardson number ( R i) which is commonly applied to dense
gas release phenomena, and represents the ratio o f the potential energies of density difference to
air/ mixture turbulence. There are several numerical definitions of Ri in use, but in this case, it
333
represents the degree of air stability within the cloud that initially develops below the release
point. A larger Ri indicates a better cloud stability, i.e., less air movement and a more defined
decent of the plume due to its higher density and/or longer fall. Thus, in isolation a smaller Ri
concurs with better mixing. Thirdly, Tty is the ratio of air discharge height to gas release height.
3.0
\0 25
\
O \ CO 20
„ 1.5
J5 1.0
-oo
0.5 H
00
-0.5
1 0.0 0.5 1.0 1.5
inn, n,
In
Figure 135: Correlation /o/TF (r2 = 0.88) Figure 136: Correlation forTli ( r = 0.97)
Correlations for the three pi-groups are given in equations (321), (322) and (323). Figure 135 and
Figure 136 show the correlations between the pi-groups for obtaining the indices and coefficients
shown in Table 46.
n, =*, •nj (321)
\n 2
n.
n , =a~, • (322)
ar n f
n3
it
111 =<*3 (323)
^ 2/(o2 .{^7] •^", })',2
Combining equations (318), (319) and (320) produced the overall correlation (equation 324)
t nr, \ ( g£ '' ■/
Kn, ^ 49 f i A163
C f.m a x ~ 0 0913 (324)
V
\ Rm J ( U Rm ' %) \ k r J
The conventional equation for Ri is employed here, but with the inclusion of the discharge angle
adjustment factor; equation (319) becomes equation (325).
334
It is noted that despite a fixed uRm, the experiments showed that Cf max changed with air
discharge angle. Because insufficient data was available to integrate this variable using
dimensional analysis, an adjustment factor ( £ ) was introduced and applied to the mean room air
speed to account for the effect of non-horizontal air discharge (equation 326). It includes the
relative height of air discharge to room height since it is expected that the influence of discharge
angle on mixing is less apparent as the height of air discharge approaches floor level, as discussed
in Section 3.1.
4 = exp [ (hd /hRm)-(a4 -0 + a5) ] (326)
The angle of the discharge, 0 , (given in radians) is defined relative to the vertical plane. For
example, 0 = 90° = n i l rad for a horizontal discharge and 0 = 60° = n 13 rad for a
downward jet. The value of the constants a4 and a5 are also included in Table 46.
Table 46: Constants coefficients and indices fo r pi-groups and for the adjustment factor
Pi-group/equation Constant Index

(321) ax =0.060 nx = -0.5799
(322) a2 =0.123 n2 =-1.72
(323) =5.611 « 3 = 1.60
(326) a4 =-1.214, a5 = 1.464 -
Substituting equation (325) into (324) and rearranging for e (as equation 318) gives equation
(327).
f u V 63
s = 0.091- • (327)
\K ;
The mean maximum floor concentration is therefore (equation 328).
C
^ f,m ax -~ F - C
b Rm (328)
where CRm is the mean room concentration ( mr IVRm). Since the experiments always showed the
mean maximum concentration to occur at floor level, C /:„!ar cannot be lower than the
homogenous concentration so equation (328) is only valid when e > 1. Lastly, it is important to
note that where no air movement is present (u Rm ~ 0) equations (327) and (328) give
unrealistically high values o f C f max . From the measurements, it was found that the air speed in a
quiescent room was consistently in the range 0.03 - 0.05 m s'1, regardless of the duration that it
was left undisturbed. The formula of Kataoka et al (2000a) was based on a quiescent room
(equation 329), so this was used to determine the effective minimum air speed.
C, ------- 1^ = (329)
-c " “
Minimum air speed was found by substituting equation (329) into equations (327) and (328) and
after putting hd = hr , rearranging for uRm (equation 330). The derived value of uRmmin
compared favourably with uRm from the measurements without forced airflow. It is also noted
that the introduction of ARm implies Cf max has a dependence on room size when forced airflow
is absent.
iW , =01 (330)
h Rm \ A Rm
When there is no forced airflow, uRm in equation (324) becomes uRmmin in equation (330).
catastrophic
slow leak
o leak
CO
hc
oo
o
o
rr(210) time
Figure 137: Gradient o f floor concentration development with time
3.2.2 Floor concentration from non-catastrophic releases
Whilst evaluation of Cf max is important for predicting the worse case (i.e., catastrophic release)
concentrations, it is also useful to predict the concentrations resulting from slower releases. This
is particularly relevant when considering the case of a release continuing after forced airflow has
ceased. Therefore, an additional set of empirical formulae was developed for leaks lasting more
than 210 s. To do this it was necessary to characterise the concentration development over time,
336
and a simple time-concentralion model was used to extrapolate concentration change over an
extended release time, as illustrated in Figure 137.
A good correlation was observed between 1!C f and time, in the form of equation (331), although
at the start of the release (typically for the first 50 - 100 s) the concentration-time profile tended
to be erratic (e.g.. Figure 106 and Figure 111), and did not necessarily match the smooth curve
fit, nor any other pattern.
a - t' (331)
CM)
where a and n are obtained from equations (332) and (333), and the subscript 1 represents the
start and subscript 2 the end of the release at Cf - C f max .
I n (l/C /tl) - l n ( l / C /f2)

(332)
lnfj - Inf2
In# = In - n - In r, (333)
Substituting equations (332) and (333) into (331) gives equation (334), characterising the
development of floor concentration with time.
1
—----- = exp In 1 1 - n • I n tra (334)
c,(f) l C/-. J
If C f max for a catastrophic leak is calculated, the increase in concentration as the leak continues
past 210 s could then be estimated, because it will be approximated by the curve prescribed by
equation (331), as seen in Figure 137. Taking equation (334), and setting C f2 = C fmax{210)
and tr2 as tr (210) for a catastrophic release, it was rearranged to find C f wfien tr > 210 s
(equation 335).
f tr (210)” ^
Cf (tr) = 1/exp In (335)
(210)
337
Based on the mass that is released after 210 s, C fmax{210) is calculated from equation (328), but
using Cfim(210) instead of CRm, as obtained from equation (336), which gives the mean room
concentration had the release stopped at 210 s.
r \
£ * .(210) = tr ( 2 1 0 ) - ^ (336)
‘r ,
vy Rm
If equation (335) is to be used for conditions other than those in the experiments, the unknown
variable n must be found (since this cannot be determined from equation (332)). The value of n
was determined for each experiment by fitting equation (334) to the experimental data (for the
latter two-thirds of the release time) until C f max is reached. The data showed that when £ » 1 ,
dCf / dt w as approximately constant, and therefore n « 1. With larger e , the curvature of
dC f / dt increased, so n < 1. This decay in the rate of increase of floor concentration is
attributed to changes in the rate of mixing near floor level as a result of the density of the mixture
at the floor approaching that of the descending plume, inhibiting the rate of mixing. From the data
set, a reasonable correlation betw een n and s was established (Figure 138), from which
equation (337) was derived.
n = -)= (337)
1.0 -1 ----------------
OS
0 .8
°o°
0 .6 - o
< OO
o
0 .4 - 9"
O
0 .2 -
yy
0 .0 - ----r ............. r -
0 .0 02 0.4 0.6 0.8 1.0
1/V7
Figure 138: Correlation between n and mixing effectiveness (r2 = 0.74)
A comparison against the measurements from tests 25 and 29 are shown in Figure 139, where the
release time exceeded 210 s. For experiment no. 25, the correlation replicates the data reasonably
well, although whilst the shape of the concentration development is comparable for experiment
no. 29, the numerical fit is poor. One observation is that for the first stage when the fan is
338
operating (to 210 s), predicted concentrations are lower than the measured. Equation (327) gives
s = 1.7, which supports the observation of good mixing in the first stage, whilst the second stage
(after 210 s) where the airflow had been terminated, the model follows the shape of the curve but
the increase in C f is over predicted. This is probably attributed to residual air momentum
following cessation of the fen producing better mixing than if the room air was “still'’ from the
point at which the fan terminated. In addition, the difference in C f during either stage could be
due to non-constant mass flow of gas during the experiment due to the fact that it was manually
modulated.
0.05
test no. 29 o test no. 25
(correlation)
0.04 a test no. 29
S 0.03
0.02
0.01
test no. 25
(correlation)
0.00
0 200 400 600 800 1000
tire (s)
Figure 139: Comparison o f calculated against measured concentrations
3.2.3 Verification of correlations

The correlations for catastrophic and non-catastrophic releases were verified. First, the accuracy
of the correlations was checked against the measurements from w hich they were initially derived.
This was achieved using all the data for AHU #1, and data for the other equipment is also shown
(Figure 140); in both sets, good overall agreement was found.
General discrepancies between predicted and measured values w as attributed to the difference
betw een o f directly measured mass of C 0 2 released and mass calculated from homogenous
concentration within room and the effect of modulating the release of C 0 2 into the room causing a
variable mass flow rate. Overall, the comparisons imply that the methods described above are
applicable to range of different units.
339
Second, in order to validate both the general correlation for catastrophic releases (equation 328)
and the correlation for non-catastrophic releases (equation 335) they were compared against data
reported by other studies. The literature review identified a number of related studies, and the
measured C fmax data from these is presented in Figure 141, against the developed correlations.
Note that in some instances, some o f the data needed for the correlations was absent, so
assumptions had to be made. Where Cf max was reported instead of C f max, the adjustment in
equation (315) was used. All of the data associated with Figure 141 is listed in Appendix 3
(Table 59, Table 60, Table 61 and Table 62).
o.io a A bki
o Impact 100 o Jabbour et al
5, 0.08 A Other units O Kataoka et al
—C bdic et al
0.06 -
lo
0.04
0.01 -
<y<>
42 0.02
0.00 0.001 —
0.00 0.02 0.04 0.06 0.08 0.10 0.001 0.01 0.1 1

Calculated ( k g m '3) C alculated C j m (kg m '3)
Figure 140: Comparison o f correlation Figure 141: Comparison o f correlations

against measurements (overall r =0.87) against data from other studies (r2 = 0.78)
The comparison showed a good match, where the average value of calculated C fmax was 16%
higher than the measured values (shown by the best-fit), and about 80% of the calculated values
were within ± 50% of measurements. These tolerances are considered to demonstrate good
accuracy of the correlations. By comparison, Cleaver et al (1994) found their model (for slow
releases in quiescent rooms) predicted mean floor concentrations to within a factor of 2 for 95%
of their experimental results. Duijm et al (1996) compared the results of two separate dispersion
models against six different sets of field and wind tunnel test data. The combined results for mean
relative bias (MRB) of maximum concentrations (considered to be a good indicator by the
authors’) show that the Britter and McQuaid models gave MRB of -0.99 to -0.55 and -0.54 to -
0.12 for the Guassian Plume Model. Using the same analysis on the current experimental data
(Figure 140) gives an MRB o f -0.007 and the validation data (Figure 141) gives an MRB of -
340
0.014. Since the optimum value is zero, these results indicate that the correlation developed here
demonstrates good accuracy.
Note that although the results from Clodic et al are presented in Figure 141. they were not
included in the best fit (or the r2 value). For this data-set, it is seen that calculated C fmax was
generally much higher than the measurements, as a results of a capillary tube being used to
release the refrigerant. The cross-sectional area of the capillary tube is small (0.004 m diameter -
compared to 0.2 m for the diffuser), leading to a very high release velocity, thus not comparable
to those used for the correlations. Further, many of the releases were made at a much lower level
(0.05 m).
Similarly, some of the tests of Aloisi employed a capillary tube, and the over-predicted data-
points are also seen in Figure 141. Examination of the data showed that at low hr , calculated
C imax was 1 0 - 2 0 times greater than measured. As hr increases, the difference reduced to a
factor of 5 at the highest release. This implies that C f max is significantly more sensitive to release
velocity as hr approaches floor level. This effect of release height and velocity was similarly
observed by Baines and Turner (1969), Marshall (1984) and Cleaver et al (1994). However,
where high airflow rate was present the differences were smaller, since the local air velocity
encourages a rate of entrainment approaching that experienced with a high initial release velocity.
Although both these data sets, based on releases from capillary tube, are inconvenient from a
validation point of view, they do provide confidence in the ‘‘safety” o f the correlations since most
releases will realistically occur as high velocity leaks, therefore designs are likely to overestimate
the “real" Cf max .
Other aspects that may have contributed to deviation between measured and calculated results
include uncertain assumptions for absent values (such as air duct outlet height and discharge
dimensions). As mentioned earlier, the correlations assume a constant release flow rate, and given
that the methods described in some of the studies suggest a variable release rate, additional
discrepancies would be expected. Additional reasons for deviation w as attributed to positioning of
sampling points, sampling intervals and o f course incorrect reporting of data (for example in
some cases, results presented C max as being less than CRm, which is a contradiction).
341
Based on the range of experimental conditions and those used in the validation studies. Table 47
lists what is considered to be the acceptable ranges for correlations, on the basis that actual
C fmax would not exceed that of equation (328).
Table 47: Range o f applicability fo r correlations
Variable Acceptable application range

Refrigerant release mass < 2 . 0 kg
Room air movement uRm= 0.02 - 0.4 m s' 1
Air discharge height 0.2 m - 3 m

Air discharge direction 7t!3 - 2 • ;r / 3 rad
Post-ventilation time 90 s - 840 s
Refrigerant release height 0.2 m - 3 m
Refrigerant release location Anywhere in room
Refrigerant release time 210 s upwards
Room size < 2 0 0 m3
Room tightness Any tightness
Room air temperature 10°C - 30°C
Room obstacles Normal furniture distribution
Release phase Liquid and vapour
Refrigerant release velocity > 0.03 m s' 1
Refrigerant release direction Any
Refrigerant density Pr > Pa
3.3 Use of correlations for design purposes

In addition to using the correlations developed in Section 3.2 for the flammable volume modelling
(Section 3 .4), it is also beneficial to employ them for the purpose of aiding design o f RAC
equipment for HCs. Although this is not strictly part o f the QRA, it is an appropriate starting
point for determination of equipment design and installation features, with the implication that
P F* is maintained at reasonably low values. Three concepts are addressed in this section:
ensuring that C f max does not exceed the LFL in the case of a catastrophic release, ensuring that
post-ventilation period is sufficient to avoid C fmax > LFL in the case of a prolonged leak, and
extending the application of the correlations to control o f airflow for the same purpose.
342
3.3.1 Maximum concentration from catastrophic leak
The first stage is to determine a maximum charge amount (or minimum room size) according to
the installation location or system needs. To be consistent with current practice, the amount
refrigerant is calculated using the approach detailed in EN 378 (equation 308), where the safety
factor q> is normally 0.2. Departure from EN 378 is possible where q> is set at a value greater
than 0.2, provided it is less than 1 otherwise mixing to below the LFL may not be possible.
Conversely, for a fixed refrigerant charge, a minimum room volume is obtained from transposing
equation (308).
The maximum tolerable concentration for design purposes C mwcdes, should not exceed the LFL.
However, to account for the regional fluctuation in floor concentrations, the LFL should be
reduced by the factor based on equation (315). So C fmax should not exceed Cmco^des (equation
338).
I FI
(338)
The variables that determine C fmax (i.e., mr , VRm, hr , hd , Va , A d , £ and p r by changing
refrigerant) were adjusted through alterations in design so that the corresponding C}max <
Cmax,des *n the case of a catastrophic release.
3.3.2 Maximum concentration from non-catastrophic leaks

For fixed conditions, a catastrophic leak is known to produce highest C f max compared to a
prolonged leak. If equipment has been designed to employ forced airflow to ensure C f max from a
catastrophic leak does not exceed C maKdes, it is necessary for the airflow' to be present for at least
the same duration of the leak (i.e., 210 s). However, if the leak time exceeds the duration of the
airflow (say > 210 s) then a C f max higher than may develop. One solution to prevent
C fmax > Cmax,des *s t° ensure constant forced airflow7(regardless o f whether the compressor is
operating, or if there is a deficit of refrigerant), but this implies increased energy use. An
alternative method is to determine a minimum post-ventilation period ( tdmm) so that C max^des is
not exceeded.
343
a
o
C3
±3
C
oo
o
e
o
W --
Figure 142: Change in concentration over time. with forced airflow ceased at td during release
To reiterate, if for a catastrophic leak, td = t r = 210 s, then C /mar is determined according to
equation (328). However, if tr > td , the additional leak time in absence of airflow may result in
an increase in rate of increase of C } with time, as illustrated in Figure 142 (also shown earlier
for tests 25 and 29). Conversely, if tr » td, then the mass flow of the release may be
sufficiently small that C f max = CRm regardless o f whether airflow' is present or not. This implies
that there is a “critical” leak time that w ill lead to a peak C fmax, corresponding to a specific td,
thus a suitable td must be determined to ensure that neither a catastrophic or prolonged release
would lead to C f max > C max^des. Characterising this trend enables the minimum post-ventilation
period ( t dmw) to be determined. In Figure 142, tdmin = 1d , which is the time corresponding to the
intercept of the two concentration curv atures that results in C mcKdes = C f max at time tr .
Assuming only two discreet stages where the first results in a concentration rise to the end o f the
post-ventilation ( ACf (td) ) and a second concentration rise to the end of release when forced
airflow is absent ( ACf {tr_d) ). Therefore, the final maximum concentration is the sum of these
(equation 339), given that
Cf , ^ , ( t r ) = b C f (td ) + LCf (tr_d) (339)
where ACf (td) and ACf (tr_d) are obtained from equation (335), and CRm is determined from
equation (336) for their respective portions of release time.
344
Since ACf {td) and ACf (tr d ) are both implicit functions of tr, equation (339) cannot be solved
directly for tdmin. Required tdmjn is when the concentration at the end of the release C f max (tr ) <
Cmax,des s according to equation (338). In practice, equation (339) is evaluated by increasing td

for a range of fr . The results of an example exercise are shown in Figure 143: when tr = t d , the
final concentration is below C maxdes, but when tr > td = 210 s, C fmox (tr ) is above C max^des. As
td was increased, C J max (tr ) falls, and in this example C f max (tr ) can no longer exceed the
Cmaxdes (of 0.04 kg m'3) provided tdmin > 510 s.
In general, a longer tdmw is needed when C f max from equation (328), when forced airflow is not
included, is significantly higher than the LFL. Conversely t dmin has a shorter duration when other
variables such as installation height, are sufficient to keep floor concentration below the LFL. It
is important for the user to conduct a proper evaluation. Firstly, if it is found that a long tdmin is
necessary’, the design and installation parameters associated with the equipment in question could
be revised, leading to a smaller tdmin. Second, a short t dndn helps minimise energy consumption
of the fan motor.
0 07
-p 0 06
B
CO
0.05
Q>
<3
O
004 - C,
tii
o
•a
e 0 03
V
210 s
0.01 t d = 330 s
t d = 450 s
000
0 200 400 600 800 1000
Release time (s)
Figure 143: Effect o f release time exceeding airflow time on fin a l floor concentration
3.3.3 Application to control of airflow

Forced airflow may be activated or terminated by pressure transducers or refrigerant detectors.
RAC equipment sometimes employs pressure transducers that upon measuring a predetermined
value may terminate the compressor, condenser and/or evaporator fans. Since a loss in refrigerant
345
charge normally results in a reduction of evaporating and condensing pressures, a certain charge
deficit may cause the pressure transducers to cease the unit forced airflow. Were the relationship
between refrigerant charge and system operation parameters known, the duration from the start of
a leak to the point at which the airflow terminates could be estimated.
Figure 144 illustrates the low rate of increase of C f due to high uRm - line (a) - followed by the
sudden increase in the C f as the controls detect the loss of refrigerant through system operation
parameters at time ( tdetect), switching off the fans thus terminating the air flow'. Conversely, a
similar control could be employed to benefit dispersion of a release, where initiation of forced
airflow is activated. This is also illustrated in Figure 144 - line (b) - where a leak initially
produced a rapid rise of Cf , until the controls detect the absence of charge at tdetect, at which
point forced airflow would be initiated thereby increasing u Rm, and subsequently lowering the
rate of C f rise, thus lowering C f max . Similarly forced airflow could be initiated on demand of a
refrigerant detector. A detector (typically located at floor level for a dense gas) would be set to
initiate emergency airflow at a predetermined concentration ( C f detect) - line (c). At the time when
first C f = C f detect, the controls would initiate airflow' resulting in lower C f max (Figure 144).
Determination of C f max for all of these schemes use that same numerical procedure as that for
evaluating the use of post-ventilation period, for example, evaluation of t detect instead of t dmm.
pressure
initiates
airflow'
detector
initiates
airflow'
f detect
pressure
terminates
airflow'
^detect
tr time
Figure 144: Change in concentration with initiation and termination o f airflow
346
3.4 Development of flammable volume/time models
To evaluate the probability of ignition (via P^Q1 and Py ) and the consequence models, it is
necessary to estimate the size and duration of the flammable volume for any set of conditions, and
to this end a dispersion model was developed.
Different options for modelling the dispersion process were discussed in Section 3.2. O f these,
CFD, ZDM and Gaussian methods are viable, and have been used for similar purposes elsewhere
(Woodward, 1998). Considering the objectives, CFD is unsuitable as it requires dedicated
software, large computing power and generally long calculation times. Gaussian models demand
extensive experimental data, and are more suited to large-scale unbounded releases. The ZDM
approach is based on discretising a space into sub-volumes and handling gas/air flows within
each, as separate, homogenously mixed enclosures, and number o f elements and calculation
complexity can be selected as desired. Given this versatility, ZDM was the preferred option.
To satisfy the needs of the probability and consequence models, the desired model outputs w ere
flammable volume, flammable mass and associated flammable times. With respect to the release
and dispersion characteristics identified in the literature (Section 1.2.3), observations from the
experimental results suggest that the design of the model account for: indoor situation only,
instantaneous and continuous finite releases, isothermal vapour only, negatively buoyant (denser
than air), low' momentum and mixing mainly driven by convective and buoyancy-induced air
flow.

The entire dispersion process was modelled by evaluating three different stages of the dispersion
process: descending plume, spreading plume and decaying cloud. Each stage was independently
discretised into spatial elements. Figure 145 shows the overall concept o f the model, where the
direction o f flow (hollow arrows) and interfacial entrainment (filled arrows) are indicated. The
plume descends from some height, entraining surrounding air until it reaches the floor (first
stage). The plume then spreads across the floor creating a layer over the entire floor (second
stage). Further addition of material from the descending plume pushes up the preceding layer,
forming a series of successive layers at certain time intervals that depend on the spreading time
(from the plume base to the wall, tsprd). Entrainment occurs between the individual layers. When
the release stops, spreading ceases and thereon the mixture is considered as a decaying cloud
347
(third stage). This sequence was evaluated from initiation of the release to the time that the
concentration within every element was below LFL.26
release source
la y e r 0
--Jr-
layer 1
sp read in g
la y e r 2 plu m e
1 an d then
- I - d ecay in g
cloud
-(flo o r) la y e r h i
des cending
plu m e
Figure 145: Diagram showing basic concept o f dispersion models
For each stage the volume was discretised so that the dimensions of each element corresponded to
a fixed time step. For the descending plume stage, the vertical dimension of the element was
based on local plume velocity and the diameter of the element varied with time due to entrained
material. With the spreading plume stage, the horizontal radial dimension was dictated by local
spreading velocity. With the decaying cloud stage that exhibits no lateral movement, the radial
dimensions of the spreading plume stage were adopted for convenience. For both the spreading
and decaying cloud stages, similarly the vertical dimension of the elements varied with time due
to entrained material. Since no new material is introduced once the release stops, the number of
layers thereon remains constant for duration of the decaying cloud stage.
For simplicity the room geometry' was assumed circular, with the release originating from the
centre of the room, moving axi-symmetrically outwards in a radial direction until the flow'
reached the wall at a distance of -yjARm / n . Converting to a circular room was considered
acceptable since both Marshall et al (1984) and Kataoka et al (2000b) found that room geometry',
whether square, rectangular, partially separated or other irregular shapes, had a minor influence
on concentration development. The size of the time steps { S t ) was governed by the duration that
26Although the value of LFL and UFL are known to vary with ambient temperature, humidity,
atmospheric pressure, and method of ignition, it is assumed that LFL and UFL are constant.
348
a concentration front took to move from the base of the descending plume to the wall ( tsprd) and
the number of spatial elements selected for the spreading plume ( N dr). Local radial spreading
velocity across the floor was assumed constant for the duration of the release (supported by floor
levels air speed measurements), hence tsprd remained constant for the release duration, as do the
radial dimension of each floor element. S t was obtained by dividing the total plume spreading
time ( t d ) by the number of radial elements ( N dr), the choice of which is addressed later
(equation 340).
(340)
As with the spreading plume, the time taken for a front to travel from the release source, down
the descending plume to the floor (called tpl) was divided by 8 1 to determine the number of
spatial elements ( N dp). tpl was calculated from the release height and local velocity of the
descending plume, which was assumed constant for the release duration. Again, because plume
centreline velocity changes w'ith distance, the vertical dimension of each element differed. At each
time step, entrainment of refrigerant and air mixture occurs between adjacent volumetric
elements. The volume flux of material transferring across element boundaries was calculated
according to the basic formula (equation 341) from entrainment theory (e.g., Baines and Turner,
1969).
(341)
dt
where A is the area of the interface and ufn is a velocity relative to the boundary, termed
friction velocity, a is the entrainment coefficient and represents the ratio of actual volume flow'
across the boundary' to the total possible flow' rate, or conventionally defined: the ratio of the
velocity of entrained material across the boundary' to the relative velocity between elements.
Evaluation of u fn and a was handled in different w ays according to the particular stage o f the
dispersion process.
349
input conditions
set airflow. A = 0 set airflow to input

conditions
calc descending and

spreading plume velocities
I calc spatial elements and

!________ time step_______
I new time step, t = t + ]
: calc descending plume and

plume spread models
a ( a c ) = f ( C Uak i C f )
does A = 0 ? ' does a c = a c{0) ?
N
_SL
calc cloud decay model
Calc flammable quantities

( V F. m - , t F)
end
Figure 146: Flow chart fo r calculation sequence offlammable quantities
The overall calculation sequence was based on an iterative procedure, as summarised in Figure
146. Based on a set of input conditions, for each stage the mean floor concentration ( C f calc) was
calculated over the duration of the release. However, for the spreading plume, calculation of u fn
and a each requires an unknown constant, o{ufrj) and a(a) respectively, to be used later. In
order to determine ct{ufri) and a (a ) , the dispersion model must be iterated for a set of
conditions where a is already established, and ufn can be subsequently obtained. Thereafter,
a{a) can be employed for the conditions where a is not known. The literature indicates a
consensus on the value of entrainment coefficient for a descending ( a p ~ 0.1) and a spreading
plume ( a c ~ 0.05) for quiescent conditions (e.g., Baines and Turner, 1969). Given the
350
descending plume and spreading velocities are known, an initial set o f calculations for floor
concentrations were made using certain selected input conditions, except uRm was set to a
minimum value according to a quiescent room.
Using an initial guess for a(ufn) and a ( a ) , the concentration of each floor element ( Cf j ) was
used to calculate the mean floor concentration at each time step (equation 342), for t - >t r .
Ndr
C f,c a lc ( 0 ~ Z Cf.,w-Af.j ' Z ^ <3 4 2 >

Vj=1 ) /=’
where A f j is the floor area covered by an element. For each S t , the result from equation (342)
was compared against the corresponding Cf as determined from the empirical formulae
(equation 335). The constant a(a) was then adjusted accordingly, and the calculation reiterated
until Cfcalc(t) = C f (t) was satisfied, when the calculation proceeded to the next time step.
Once the final calculation for t = tr was complete, the time-averaged a c for the spreading plume
was compared against the established value (~ 0.05) for quiescent conditions. a{u fri) was then
obtained accordingly and the entire set of calculations reiterated, until the mean calculated a c
was satisfied. a(u fn) was then fixed for the final iteration of the entire model with the
introduction of forced airflow7conditions. Subsequently, empirical C f (t), and Cfcalc(t) were re
calculated for the new uRm, which requires only one iteration since the constant a(u fri) w as now
fixed.
For the decaying cloud, a c was similarly used to calculate Cf calc{t) . However, it was not
possible to determine a{ac) according to Cf (t) since the empirical correlation is not applicable
once the release has stopped. Instead, a ( a c) was fixed at the value determined for the final time
step of the spreading plume (i.e., at tr ), and used for the duration of the decaying cloud stage.
The calculation sequence ceased once the concentration within every element was below LFL.
Recalling the purpose of the model, values taken during the final iteration from each time step
when Cf calc(t) = Cf {t) were used to calculate the various flammable quantities. Within each
room reference volume (corresponding to the three room height elements, h ) the flammable
351
volume-time, (V Ft F )h, and the flammable mass-time, (m Ft F) h, are calculated by summing the
flammable volume of all the spatial elements, j , at each time step, i , until all the mixture is
below the LFL (equations 343 and 344, respectively).
(343)
(344)
where v £ t is the exit volume of each flammable element whether in the descending plume.
spreading plume or deca\ing cloud, and C Fut, is its corresponding concentration. The total
flammable volume-time and flammable mass-time are the addition of all three room height
elements (equation 345 and 346, respectively).
(345)
(346)
h=\
Lastly, the mean flammable volume ( V F , equation 347), and mean flammable mass of the
refrigerant ( m F , equation 348), are determined by dividing the totals from equation (345) and
(346) by the total flammable time, t F .
(347)
(348)
n—i
Summing the refrigerant mass within all elements (as VoutJ x Cout J) gives the total mass released
up to a time t provided infiltration and extract ventilation are absent.
3.4.2 Spreading velocity

It is important to determine the spreading velocity of the plume, as it is needed to calculate the
dimensions of spatial elements, and because it influences entrainment in that it is needed for
352
calculation of the friction velocity. From the literature, two formulas were found for estimating
velocity of the flow o f a gas release across a flat surface. CCPS (1996) details an approach of
Fay, which is similar to that of Sykes et al (1999) and when transposed, gives equation (349).
“ sprdj = { 4 - g ' - m r ' t sprd / Pr 4 'tj (349)
Alternatively, Ditali et al (2000) give a formula that is also transposed for velocity (equation
350).
7 g ’- i hr' t )
u L (350)
2 p r -r]
where r is the radial distance from the release source at time t .
When equations (349) and (350) were compared against the test data, it was found that there was
a poor correlation (Figure 147), possibly due to not accounting for release height, air movement,
etc. In addition, as Sykes et al (1999) point out, the velocity of a spreading plume varies with the
height of the layer above floor level, which implies it is important to specify spatial location of
the velocity'.
0.5
10
0.4
1-
3 0.3
0.2
0.1 -
JS * x o correlation
3
JJ - Ditali etal
a x *
U x CCPS
0.01
0 0.1 0.2 0.3 0.4 0.5 l.E+3 l.E+4 l.E+5 l.E+6
Measured ( m s 1) Reynolds number (-)
Figure 147: Comparison ofplume spread Figure 148: Correlation o f friction factor and
correlations and measurements (r2 = 0.43) Reynolds no. fo r spread velocity (r2 = 0.67
To address these points, a new correlation was developed based on Bernoulli. To include the
influence of the release characteristics, the room air speed should also be considered since it may
act against or contribute to the momentum of the moving front. The new approach addressed the
353
static, velocity and frictional forces acting on the moving front. First, the static pressure imposed
by the front was calculated from equation (351).
APstaJ = g hef f - ( Po~Pa) (351)
where p 0 is the mean density o f the mixture at the base of the plume; this is used instead of local
density because it avoids excessive iterations o f the entire model to obtain the local concentration.
The effective cloud height heff is based on an ideal homogenised cloud at C fmax (equation 352).
mJC<
V = (352)
ARm
Second, the velocity pressure was approximated according to the difference between the plume
front and the opposing force imposed by the room air (equation 353).
APvelj = \ (Po *U lp r d J - Pa ’ ) (353)
Thirdly, the frictional pressure was obtained from the conventional formula for flow across a flat
surface (Douglas et al, 1986), equation (354).
APfiij = 8 • V Ai • ulPrd.j (354)
where the friction factor Xj was obtained from (equation 355).
X j = c R e n} (355)
where the Reynolds number Re j was calculated according to a modified version of the
conventional formula (equation 356).
„ Po ’ Msp rd ,j I ~> . ->

Re = --------------- J r f ■ +h ; (356)
p v
where p is viscosity of the mixture and was approximated to 0.00002 Pa s. The length
modification ( j + h; ) was included because a much closer correlation was achieved with the
direct distance from the release source height, instead of simply using rf i . Finally, the local
spreading velocity uswd j was obtained by iterating through equations (351) to (356) until
equation (357) was satisfied.
4P fn j + *Psta.j + 4 ? V j = 0 (357)
354
The unknowns in equation (355) were found by applying the data from the experiments in Section
3.3.1 to equations (354) to (356). The data consisted of a combination of both floor level
anemometer measurements for the duration o f the release, and time differences betw een initial
concentration measurements from different floor sampling points following the first positive
concentration measurement directly below the release point. Both data sets provided
approximate!} similar results. The correlation between X and Re is shown in Figure 148, and
for equation (355) the best-fit gave c = 37570, and n - -1.1. Typically for flow7over a flat
surface, -1 < n < 0 (Douglas et al, 1986). The scatter was attributed to the measurements being
of inadequate quality'. Since the experiments were not specifically designed for obtaining the
plume spreading velocity-, it is believed that errors were introduced by insufficient frequency of
sampling times, anemometers providing unidirectional air speeds and the influence of near walls
on flow patterns. Nevertheless, the proposed approach is better than the formulae found in the
literature, and therefore it was employed in the dispersion model. Further, the new approach can
be developed when more reliable data becomes available.
3.4.3 Descending plume model

When an aperture develops within the refrigerant circuit, the leak is assumed to be released at a
constant mass flow rate that is determined based on the released mass and the leak time (equation
358).
mr = — (358)
^r
where is the design charge size o f one circuit o f the RAC equipment, i// is a refrigerant
retention factor (Section 2.4), and tr = tleak .
The leak produces a high velocity7jet, which may impinge on a surface or flow directly into an
open space in any direction. Ultimately, whether through impingement or entrainment, momentum
gives way to buoyancy forces and a relatively low velocity plume is formed that falls towards the
floor. As it travels downwards, it entrains more air from the ambient thereby increasing in volume
flow rate ( Vp out > Vp m) and decreasing mean radial concentration ( C p ), density and centreline
velocity7( u p ). In addition, some mixture is entrained out of the plume into the surroundings. This
process can be follow'ed in Figure 149. The basis of the plume descent model uses the
355
fundamental formula for evolution of a plume in a quiescent environment, as developed by Baines
and Turner (1969) (equation 359).27
dI l± = ^ D r a f - umy S z l (359)
where subscript e denotes entrainment and D ■is the local plume diameter.
Vr release source
entrainment out
Rm
Sz
entrainment in
p .o u i
descent
distance, z floor
Figure 149: Schematic diagram o f plume model
For a circular plume with similar density7to the (quiescent) surroundings, a p was reported to be
constant at around 0.08 to 0.10 (Baines and Turner, 1969), where the plume centreline velocity
u p was used as u fn . CCPS (1996) gives a p = 0.9 for a high density gas release ffom an
outdoor vent stack, whilst Thomas (2000) stated that a p is not constant but dependent upon
plume density (although no values or correlations were provided). In the current study it is
necessary to account for the influence of forced airflow within the room and therefore - consistent
with the approach of Etheridge and Sandberg (1996) - a p was fixed (at 0.10) and u fn j was
calculated according to relative velocities of the plume centreline and the cross-wind (i.e., uRm),
termed the “excess velocity”. Conventionally, the friction velocity (equation 360) is a function of
the fluid boundary7friction stress (equation 361).
ufii =^ Tfn/ P <36°)
27 Zhao et al (2002) presented a more detailed Gaussian approach to determining concentration profiles
within a jet. It was not employed here because it assumed only a high velocity jet and could therefore
lead to an underestimation of concentrations.
356
T^=M-— (361)
* M dh
where duldh is the velocity gradient with respect to distance ( h ) from a boundary. By
substituting equation (360) into equation (361), the friction velocity was obtained (equation 362).
(362)
where a(u ^ ) was included to account for effects of turbulence as a result of the absence of a
solid boundary, and was determined through an iterative process discussed in Section 3.4.1
(Figure 146). d uldh was based on the combined local plume centreline velocity and the mean
room airspeed (i.e., up . + uRm) as well as twice the radial distance from the centreline (i.e., 2 *
Dj 12). up j was calculated using the solution derived by Etheridge and Sandberg (1996)
(equation 363).
(g'-v )1/3
u p,j = 0.437 •— r \ ri (363)
a p ’z i
where Vr is the release volume flow rate of the refrigerant at room temperature and atmospheric
pressure (equation 364).
Vr = m r ! p r (364)
The local plume diameter was equation (365).
Dj =2.69 -ctp-Zj (365)
Equation (359) assumed that entrainment occurs only into the plume, whereas it is known from
both initial visualisation tests and from concentration measurements (Section 3.1) that material
from the plume is also entrained into room air. This is handled with a gas exchange coefficient
(/? ) in equations (366) and (367) for volume flow7, and equations (368) and (369) are
corresponding mass flow of refrigerant into, and out o f the plume respectively.
dV . . dVej
(3 6 6 >
= (367)
dt dt
357
d m r e '» J _ c . d V e i n_ il. (3 6 8 )
dt surr'1 dt
d m r ,e,out,j _ p r ^ e,o u t,j (3 6 9 )
dt “ °UtJ' dt
where - is the mean concentration within the local surroundings, and CoutJ is the mean
concentration at the outlet of an element (equation 370).
^ o u t , j = ™ r,o u t,j I V o m j (3 7 ° )
The total volume and mass balance for each element is equation (371) and (372) respectively.
d V o u t.j = d y in,j | dVeJnJ dVe out J

dt dt + dt dt
d m r .o u ,J _ d m r ,,n j | d m r ,e ,in ,J (3 ?2 )
dt dt dt dt
Equation (341) states that entrainment is proportional to the velocity difference between the
plume centreline and the environment. Therefore, f$ was simply assumed proportional to the
relative velocities between the plume centre line velocity and the room mean airspeed, as
expressed by equation (373).
J3} = Uj_ (373)

U j + U Rm
Given that plume centreline velocity7is normally an order of magnitude greater than uRm, the
majority of material is entrained into the plume as seen in equations (366)and (367).
The output from equations (371)and (372) provide values of volume and mass forinclusion in
the calculation of V F (equation 347) and (equation 348).
3.4.4 Spreading plume model

The work of Baines and Turner (1969), Cleaver et al (1994) and others showed that when a
plume within a quiescent room reaches floor level, the flow spreads laterally; further entraining
air until a layer is created. Depending upon the local momentum and buoyancy, the layer can be
fully mixed (e.g., when Ri < 10) becoming homogenous, otherwise a second stratified layer that
graduates from the concentration of the homogenous layer to zero, is created above it. With
358
forced airflow, similar two-layer patterns are demonstrated experimentally and in various studies
detailed in the literature.
Based on these characteristics, the plume spread model was based on a multi-layer approach
where an initial floor layer is formed. As it spreads, the flow entrains material from above and
subsequently dilutes. As the descending plume continues to the floor, the new denser front
displaces the existing diluted mixture thereby creating an upper layer. This process continues
until the release ceases and no new layers are formed. Each successive layer entrains material
from the layer above it.
As with the descending plume, entrainment theory was also used to calculate exchange of
material across the interface of tw o layers, although adapted in respect of the stability o f the
lower layer. The properties of the lower layer were accounted for with the inclusion of the local
Richardson number to produce the conventional formula for entrainment of dense clouds (e.g.,
Britter, 1998) (equation 374).
dVa . A , • , -u ,
— *_± = _ j AC — tz L (3 7 4 )
dt RiCJ
where A- is the area of the interface between the two layers. The review of Kumar et al (2003)
identified from most previous studies on cloud dispersion that entrainment was dominated by the
horizontal interfaces, so entrainment into any “vertical” moving front (when existing) was
neglected. The local Richardson number ( RiCJ ) is defined in equation (375).
n c j = ^ Lr - L (375>
ufiij
where g ' is based on the local density. The local friction velocity ufn j is handled in a similar
manner as with the descending plume. Following the discussion for descending plume
calculations, u fri J for the lower/upper layer interface also employed equation (362), where the
velocity gradient ( d u ld h ) is the difference between layer centre velocities with respect to
distance (height) across the boundary (from equation 376). The local centreline velocity7was
based on interpolation between u sprd j ■

>uRm and zero velocity at the ceiling. However, since the
spreading plume flow was assumed to be counter to the room airflow (due to return airflow back
to the AHU), a combined velocity is employed (u Rm + u d ■). With this, interpolation o f velocity
359
relative to height was achieved by using a simple power-law equation (376) (as for wind speed
relative to height).
u(h) = a .( h Rm- h r (376)
where a and n are determined from u - 0 at h - hRm, u * uRm at h = hd, and u = uRm + usprd ]
at h = 0 , as illustrated in Figure 150. Although using hd as a reference height implies that the air
discharge centreline should be the corresponding velocity, such high air speeds would only be
present within a localised region; uRm is considered to represent the average lateral air speed. In
the case of a non-horizontal directed discharge duct, h is calculated as the mean height of the
discharge centreline across the room - or the average height of the lowest layer h}, w hichever is
the highest.
A similar approach was validated experimentally by Brown et al (1997) for unbounded (outside)
releases. It is known that the density of layers influence the velocity profile and this is handled
comprehensively by Sykes et al (1999). The effect was seen to be minor, so it has been neglected
here.
ceiling height £
air discharge height
plume spread height
Figure 150: Illustration o f room velocity profile
Evaluation o f entrainment coefficient for a spreading plume or cloud ( a c ) within the context of
transient or “static” stratified layers formed by a small release in a confined space w7as not found
in the literature; most addressed large-scale outside releases. For the surface of a dense cloud,
Gopalakrishnan and Sharan (1997) reported a c to be proportional to the change in cloud mean
density as it disperses over time. Tickle (1996) considered dispersion of a high density gas
360
travelling down a slope, and by correlating a model with measurements found a c ranged from
0.026 - 0.045. Baines and Turner (1969) and Etheridge and Sandberg (1988) proposed a c =
0.04 to 0.05 for a plane or spreading plume under quiescent conditions. Britter (1988)
summarised various correlations for a c , again determined from large outdoor releases. These
were considered for use in the model and their generalised forms are presented as equations (377)
and (378).
a c = a { a c)-Ricx (377)
a c = a (a c)/(l + b -R ic) (378)
As cited by Britter (1988), for equation (377) Kanatha et al gives a (a c) = 5, from Lofquist
a (a c) = 1.4, and McQuaid, a (a c) = 2.85 (although in calculating Ric , McQuaid has the
reduced gravity as a function of mean density, g ' = g (p / p a - 1), instead of the conventional
formulation). For equation (378), Havens and Spicer give a (a c) = 0.4, and b = 0.125. The
various constants were derived under differing ranges of Ric and for different release conditions,
such as outdoor ground source plume and outdoor elevated plumes; none of these wêre derived
for the present indoor release case, so each were assessed within the model. Subsequently,
equation (378) was selected, but a (a c) was adjusted to suit the empirical data for the indoor
case. Lastly, equation (377) produced instability within the overall model when applied to the
wide range of Ric values normally experienced.
The volume and mass balances for each element within each layer was handled consistent with
the descending plume, so equations (366), (367), (368) and (369) also apply for the spreading
plume, as with the calculation of the exchange parameter, /? . In equation (366), Csurrj
represents the concentration within the element above the interface. However, because the
presence of mechanical (extract) ventilation or infiltration may remove material from the room an
additional term is required to account for an overall loss of refrigerant mass. Assuming that room
internal pressure is maintained, the volume of air removed will be replenished (with air containing
no refrigerant), so the volume based equation (371) still holds. On the other hand, as air is
removed from the room there is a net loss of refrigerant mass, so equation (372) w as adapted
accordingly (equation 379).
(379)
dt dt dt dt
361
where dmr vk / dt is the mass flow of the refrigerant out of each element due to the particular
type of ventilation (where the subscript k corresponds to the types in Section 2 .1), denoted by
(equation 380).
dmr vk — dl\. t
— = Cft»('>---- — <38°)
dt Rm dt
In using the term for mean room concentration, equation (380) assumes that the ventilation
removes a homogenous gas/air mixture; an approach adopted here for simplicity7, coj k in
equation (379) represents the mass fraction of material removed from each element, and is simply
assumed to correspond to the mass fraction of the given element to the entire content of
refrigerant within the room at any7given time (equation 381).
cok j (r) = mr ] (t) / £ m r j (t) (381)

j
The output from equations (371) and (379) provide values of volume and mass for inclusion in
the calculation of V F (equation 347) and w F (equation 348).
3.4.5 Cloud decay model

At the end of the release, if any Cou! j are above the LFL then the decaying cloud model was
employed, since a flammable volume still exists. The cloud decay model was similar to that for
the spreading plume in terms of discretisation and entrainment equations. At cessation o f the
release, the descending plume w as assumed to '‘disappear'’, and the remaining layers of the
spreading plume become ‘'static”, and no new layers are formed. Although there is no longer
horizontal flow across the floor, a velocity7gradient still exists due to uRm , so entrainment
continues. The rate of entrainment is again calculated with the appropriate equations (366, 367,
368 and 369), although adapted in two ways. First the velocity profile is re-evaluated for u f ] =
0, since lateral movement of the mixture is assumed to have ceased. Second, as mentioned in
Section 3.4.1, it was not possible to calculate C f (t) from the correlations, so calculation of
C f^caic (f) required a (a c) to be fixed at the value determined for the final time step of the
spreading plume (i.e., at tr ). This produced comparable results to the approach of
Gopalakrishnan and Sharan (1997), where a c is extrapolated over time according to the change
in local density. Since horizontal flow7within the cloud has ceased, each parcel remains at the
same lateral position as from time t r . The calculation sequence continued until all Cout . < LFL.
362
The main output from the cloud decay model is calculation of volume and mass within the
elements - again using equations (371) and (379) - for inclusion in general equations for
determining V F (equation 347) and m F (equation 348), except where C f calc is based on fixed
a{a c) , and no further refrigerant is added to the layers.
3.4.6 Airflow models

Characterising room airflow is important for dispersion modelling, and this section summarises
determination of uRm for the different forms of airflow. From Section 2.1, the four sources of
airflow under consideration are: unit airflow, mechanical ventilation, infiltration, and thermal
convection. Whilst airflow from infiltration and thermal sources are independent of the operation
of the refrigerating system, unit airflow and mechanical ventilation can be linked to the
occurrence of a leakage. Modelling of airflows within rooms can be very complex, but for the
purposes of this study, a relatively simplistic approach was be taken, and models that are more
complicated can be applied at a later stage if necessary. In addition to the airflow7types listed
above, movement of people within an environment increase air speeds, but because of the
complexity of this problem, it was neglected.
Unit airflow
For airflow from RAC equipment, volume flow7rates are specific to the equipment being
considered. Generally, airflow rate is a function of the thermal load ( 0 ) , typically Vunit ~
0.03 Q to 0 .3 -0 , depending upon its function, as for an evaporator or condenser. Reference to
Table 40 provides some examples of selected equipment.
Mechanical ventilation
Mechanical airflow rate is governed by the application and may be used for a number of purposes
including heat and odour removal, or extraction of leaked refrigerant. (Note that calculation of
uRm for mechanical ventilation using extract vent only w as treated as if it were infiltration
because air speed is dictated by the room fabric leakage, and not on inlet vent area.) Air speed
from mechanical ventilation that employs a specific inlet duct was calculated as with unit airflow
and A v is the duct area (rather than the outlet duct). Specifically, mechanical ventilation for
buildings other than dwellings requires a minimum airflow, which is 5 1 s'1per person or 3 1 s'1
per m2 in the UK (BS 5925: 1991). Alternatively, refrigeration safety standards provide options
363
for machinery rooms, including a flow rate based on refrigerant charge ( Vmech = 0.014 • m 2r ") or
a nominal A mech = 10 h'1 (EN 378: 2000).
Infiltration
Infiltration occurs in all buildings due to air leakage through fabric, caused by an imposed
pressure difference ( Ap ) between the internal and external environment, and calculation of
infiltration rate was detailed in Section 2.5. For an imposed Aptot other than the standard
reference Aptot = 50 Pa, infiltration rate was calculated with equation (382). Note that A 50 is
halved on the assumption that inflow occurs through half of the room fabric surface, instead of
the total surface applicable to when the room is pressurised to determine A 50.
v0.7
v* infil -- ^ ~50 JSptot
(382)
50 J 3600
Combining equation (382) with the steady flow equation, the room air speed due to infiltration is
obtained (equation 383).
(2 -5 0 I p )1/4
■«. = v •* ■V(A5<
/.. 1-3 6 0 0 /2 )/c
. v „.
, -V'r
<383>
Rm
where the discharge coefficient cd is 0.6. For infiltration, a survey by Orme et al (1998) provided
A 50 data for dwellings and non-domestic buildings in several countries. In the UK A 50 ranged
from 3 h'1to 30 h 1, giving a statistical mean value o f 14 h"1. Using equation (383) and the data
within the example calculations in Section 2.5 (Figure 101), the range of AAjr[fil for the specified
room produce uRm of 0.015 to 0.03 m s'1 (for infiltration only).
Thermal convection
To evaluate airflow due to thermal sources, an approach from Etheridge and Sandberg (1996)
was adopted. Assuming an equal heat output per object, the equivalent volume flow rate of warm
air per object is calculated from equation (384).
= 2 - 2 3 - V J ) 5' 3 -(2 .8 x l0 -5 - 0 1' 3 (384)
where Q is the average heat input into the room per object, htherm is the height o f the object, and
a =0.083. To calculate uRm the discharge area in equation (317) is substituted by the effective
area of all thermal sources: N therm x A eff, where N therm is the number of items. The value of
364
Aeff was found to be 0.56 m2 per object based on the measured air velocities in test no. 43. In
practice, the number of thermal sources depends on the occupants and purpose of the room (e.g.,
an office with occupants, PCs, monitors and other electronic equipment).
Multiple sources of airflow

Where airflows originate from different sources simultaneously, RMS values of uRm were used
as Etheridge and Sandberg (1996) (equation 385). Also the correlations in Section 3.3 require a
corresponding height for the discharged air, a uRm-weighted RMS was used to determine
equivalent hd (equation 386).
«*. = I Z s L j, <385>
hj - ^ URm7 (386)
V k k
3.4.7 Model verification

Two separate exercises were carried out to verify the dispersion model. These were evaluating
sensitivity to the number of elements and consistency with measurements.
Table 48: Input details fo r dependence o f flammable volume-time on number o f elements
Run number 1 2 3 4 5 6
Release mass (kg) 0.50 1.50 1.50 1.50 0.50 1.50
Room area (nr) 20 45 45 45 20 45
Room height (m) 2.5 3 3 3 2.5 3
Leak time (s) 240 240 240 500 500 500
Release height (m) 1.5 1.5 1.5 0.5 0.5 2
Unit height (m) 1.5 1.5 2.5 2 2 2
Airflow7(m3 s'1) 0.014 0.038 0.113 0.131 0.049 0.019
The number and dimensions of the elements within the respective sub-models was dictated by the
choice of radial elements, N dr. Conventionally, the preferred number of elements is obtained by
comparing outputs for an increasing number of elements until the successive change is negligible.
For this exercise six different sets of input conditions were chosen (listed in Table 48, plus
365
general inputs: R290, airflow only from horizontal discharge angle, 0.12 m: duct area). From
these, a range of total flammable volume-times ( V Ft F ) was generated for N dr from 10 to 250.
Figure 151 shows the results for the six different conditions, as a function of increasing N dr. In
general, V Ft F changed erratically - up to about ± 20% of the average value - until N dr - 5 0 .
By N dr ~ 100, the variation in V Ft F had smoothed and after N dr ~ 150, V Ft F remained
approximately constant. Between N dr ~ 150 to 250, the deviation was typically within ± 1% of
the final values, although a little higher for run 6, (±3% ). Compromising calculation time and
desired precision N dr =150 was considered appropriate.
2500
Run 6
4 iiW,
3 2000
Run 2
& 1500
J3
"o Run 3
>
1000 ^ Run 4
Run 5
I 500 Run 1
50 100 150 200 250

N,
Figure 151: Dependence o f number o f radial elements on flammable volume-time
Consistency of the output with measurements was assessed by comparing concentration-height

profiles. This is relevant since flammable volume has a strong dependence on the vertical
concentration profile. Comparing calculated and measured Cf (t) was not necessary' because the
model was already calibrated against the empirical correlations. Rather than checking against the
entire set of experiments, five were selected to represent a wide range of conditions:
- No airflow (Figure 152) - Test no. 9
- Low airflow (Figure 153) - Test no. 28
- Reference case (medium airflow) (Figure 154) - Test no. 11
- Low release height (Figure 155) - Test no. 20
- Low release and air discharge height (Figure 156) - Test no. 21
366
3 3
120 s 120 s
240 s 210 s
2 2
p
(measured - solid) (measured - solid)
oo (model - dashed) i (model - dashed)
1 1
0 0
0 0.01 0.02 0.03 0.04 0 0.01 0.02 0.03 0.04 0.05
Concentration (kg m-3) Concertration (kg m-3)
Figure 152: Comparison ofheight- Figure 153: Comparison ofheight-

concentration profile at 120 s interval from concentration profile at 120 s interval from l
model and Test no. 9 (n o airflow) model and Test no. 28 (low airflow)
The results of the comparisons are shown as concentration-height profiles at 120 s and 210 s
(240 s for test no. 9) from the start of the release. The results from the model are shown as
dashed lines, and the measurements are solid lines.
Overall, calculated concentration-height profiles were generally consistent with measurements,

particularly at the floor and 1 m. The greatest differences occurred at the ceiling, where the
predicted concentrations were always around half the measured values, although offset by slightly
higher floor concentrations. This was possibly because the model does not directly account for
entrainment of the release into the return airflow7, which is then discharged to high level.
However, because there was only one sampling point at 3 m, the measurement may be
misrepresenting the average concentration at that height. Given this uncertainty, the deviation in
concentrations at 3 m was not considered a significant problem. In addition, the concentrations at
ceiling height were always well below the LFL, so C F would not occur in that region anyway.
367
3
2
B I (m easu red - so lid ) !
JS
W ) (m o d el - d ash ed ) ;
'5
PC
1
0
0 0.005 0.01
Concentration (kg m-3)
Figure 154: Comparison o f concentrations fo r reference case (Test no. 10)
Calculated (or interpolated) concentrations at 1 m were sometimes lower than the measurements
within the initial period of the release. Tow ards the latter part of the release, calculated and
measured concentrations converge. In general, these comparisons provide confidence in the
accuracy of the model.
120 s / 120 s
i/
2 2
j -<>—210 s
t
(m e asu re d - so lid ) ; (m e asu re d - solid)
(m o d el - d ash ed ) : (m o d el - d ash ed )
0 j o—
0 0.02 0.04 0.06 0.005 0.01 0.015

Concentration (kg m-3) Concentration (kg m-3)
Figure 155: Comparison o f concentrations fo r Figure 156: Comparison o f concentrations fo r

low release height (Test no. 20) low release and air disch. height (Test no. 21)
3.4.8 Application
The model w as used to provide an insight into the dispersion process, and how certain variables
affect its output. The results were used initially to describe the behaviour of a typical release, and
secondly to illustrate the effect of selected variables on V Ft F . The main variables that were
considered are: mean room air speed, released mass, release time, post ventilation time, room
area, release height, and air discharge height.
368
Behaviour of modelled dispersion
The behaviour of the dispersion o f a release was analysed with an example - a 1 kg release of
R290 over 250 s from a height of 2 m into a 3 m * 40 m2 room, with an airflow rate of 2 h'1room
air changes (giving uRm = 0.08 m s'1), also discharged from 2 m in a horizontal direction.
120
100
G
0.06 - o
00 £
o
>
I 0 .0 1 - <u
>
91 0
E ifil1/*tif <11
i'ii
U 0.02 - layer 11
0 100 200 300 400 500 0 100 200 300 400 500
Time(s) True (s)
Figure 157: Concentration within individual Figure 158: Cumulative volume o f individual
layers layers
Figure 157 shows how concentrations vary between layers over time. The square data points
illustrate the development of C f up to Cf max at each time interval (i.e., tsprd) for each layer, and
how7the concentration in the layer, say layer (10) reduces with time as weaker mixture from the
layer above, say layer (9) is entrained. It is seen that at the start o f the release, the material within
the layers tended to mix rapidly within about 30 s, taking up to about 60 s towards the latter part
of the release due to a lessening density7difference between the descending and spreading plume.
Dilution of the final two layers (10 and 11) created toward the end of the release occurred less
rapidly and concentration tended to decay slowly over time. This was because o f floor level
spreading velocity becoming absent, causing a drop in friction velocity7.
Also in Figure 157 is the mean concentration of the flammable elements - ( C F = m F ( t ) I V F (t)
- which falls from about stoichiometric concentration (-0 .0 7 kg m°) towards the LFL. The
downward trend of C F is expected since more of the elements approach the LFL over time,
whilst accumulating material to produce a larger volume. Throughout the flammable time, the
369
proportions of flammable to total mixture volume and the flammable mass to release mass both
vary in a similar manner as C F .
Figure 158 illustrates the cumulative change in layer volume, where the upper line (1) represents
the total volume of the entire refrigerant/air mixture. It is seen that as each layer was created, the
volume of the layer above it approaches zero, as its contents were gradually entrained into the
next layers beneath. This entrainment produced dilution in successive layers as described for
Figure 157. Once the release stopped the existing layers cease to be entrained into lower layers
and instead increase in volume as a result of continued entrainment from the upper part of the
room. Eventually layer (11) will completely fill the room with the mixture, as its concentration
approaches CRm.
Influence of selected variables

The influence of certain variables associated with the design and installation of equipment on
V Ft F was addressed, and calculations were based on a set of reference conditions, similar but
not identical to the reference test no. 10: m r = 1 .0 kg, ARm = 40 m2, hRm = 3 m, hr = 2 m, hd
- 2 m, Ad = 0.115 m2, A = 1 h'1, 6 = n i l , td = q o , and the refrigerant w7as R290. It is noted
that low7airflow was selected so as to emphasise the effects of other variables that may not
otherwise create a noticeable change in V F t F .
1E+06 T 1E+06
“ 1E+04 & lE t0 4
a
1E+02 - «C“ 1E+02 -

= 40 m2 mr =i.o kg
Cjh, =0.008 kg m-3 On,,, = 0.008 kg m-3
1E+00 1E+00 n--------------r -i-------- !
0 0.5 1 1.5 2 2.5 20 40 60 80 100
Released mass (leg) Room area (m2)
Figure 159: Effect o f released mass on Figure 160: Effect o f room area on V 7f
Figure 159 shows the influence of released mass on V F/ F from two respects: where the room
area was constant whilst mass was varied, and where room area w as simultaneously increased
370
with mass so that CRm is constant (whilst adjusting Ad to maintain uRm). With constant room
area, an increase in mass produced an almost exponentially larger V Ft F , although it was not
until mr >0.8 kg that the gradient changed and much greater values were achieved; the change
of gradient coincided with C f max > LFL. On the other hand, if room area is correspondingly
increased, the change in V Ft F was almost linear with mass and the transition o f Cfmax to above
the LFL (0.5 kg) does not demonstrate the change in gradient seen with fixed room area.
Similarly, the effect of changing room area (and Ad ) with fixed mass is presented in Figure 160,
and included again for comparative purposes is the data for constant CRm. Enlarging room area
causes V Ft F to decrease exponentially until about 50 m2. At this point, the gradient changed and
V Ft F became relatively constant with additional area. Again, this transition occurs when Cfmax
~ LFL. As with the previous example, maintaining fixed CRm by simultaneously increasing mass
produced a linear rise in V Ft F with room area.
1E+06 t 1E+05 r~
td = 210 s
1E+04 -
V3 td =315 s
1E+04 0 = 2 *.
r*“>
g B
1E+03 -
k,
C* 1E+02
^ 1E+02 -
1E+00 1E+01
0.05 0.1 0.15 0 200 400 600 800
uRm (mS" ) Release t h e (s)
Figure 161: Effect o f room air speed and Figure 162: Effect o f release time and post
discharge angle on ventilation period on
Air movement is created by the release displacing air, and forced airflow' from the AHU.
Neglecting the contribution of the release, the influence of unit airflow - in terms of uRm - on
V Ft F was addressed (Figure 161). Also included was data for an upward and downward
directed air discharge. For horizontal air discharge ( 6 = n /2 ), there was a rapid reduction in
V Ft F until uRm reaches approximately 0.05 m s'1, at which point a more gradual fall occurs.
After about 0.1 m s'1, the effect of air speed seems to have negligible impact on V Ft F . This was
consistent with the behaviour of Cf max found in the experiments. Where the air discharge was
directed downwards ( 9 = n ! 3). V Ft F was distinctly lower, particularly at lower uRm . Again,
the same trend was observed, where the reduction in V Ft F with uRm became more gradual, and
then almost constant above uRm = 0.06 m s'1. Similarly, an upward directed discharge also
produced similar behaviour but required higher uRm to achieve the same effect. In all three cases,
the initial transition from high to low V Ft F occurred when Cf max ~ LFL. The second transition
concurred with the mean concentration at the base of the plume falling below7LFL. Thus the
entire flammable volume occurred only within the plume, and thereon higher uRm had little
influence on V Ft F .
V Ft F for a range of release times (210 s upwards) is shown in Figure 162, where airflow was
continuous (td = x). In addition, results were also generated for three different post-ventilation
periods (which could also be interpreted as time for airflow7termination due to equipment
controls). With continuous airflow7, increasing t r showed V Ft F to fall exponentially until tr >
300 s, when longer release times had no further effect. Termination airflow at 315 s and 210 s
caused a minor increase of V Ft F with a 210 s release, because the flammable cloud persisted for
longer. But with longer t r , the difference disappeared because uRm has a lesser influence on
dispersion. For td - 105 s, a distinct rise in V Fl F was observed because forced airflow was
absent for half of the release, so thereon only minimal uRm was present to disperse the mixture to
below LFL. For longer tr , with Cfmax < LFL, so the td ceased to influence V Ft F .
Figure 163 illustrates changes in V Ft F with release height (hollow7data-points), but also where
air discharge height was equal to the release height ( hd = hr , filled data-points). To elaborate on
the effect of these parameters, additional data for a higher airflow was included. For hd = 2 m,
low ering hr produced a steady rise in V Ft F , which applied to both uRm , although o f course
V Ft F was always smaller for the higher uRm . As found previously, the gradient of V Ft F
flattens as Cf max < LFL, and when uRm - 0.08 m s'1, a second step change was observed
corresponding to the mean plume concentration at floor level being < LFL. When hd = hr , a
372
different tendency was observed; for uRm = 0.03 m s'1 a small V Ft F occurs at low hd and hr ,
but elevating hd and hr increases V Ft F until about 2 m where above this V Ft F began to
reduce. At low hd and hr , the high local airspeed caused rapid mixing within the developing
cloud near the floor thereby offsetting the high Cf max that would normally be associated with a
low level release. As hd and hr were raised the high discharge velocity became less effectual.
Above 2 m the release occurred from a sufficiently high level to offset the absence of airflow near
to the floor resulting in a fall in V Ft F . With uRm = 0.08 m s '\ the airflow was always sufficient
to ensure C fmax < LFL, so V Ft F remained small regardless of hd and hr .
1E+06 — _ 2m
uRm = 0.03 m s
uRm = 0.08 m s
“ 1E+04
1E+02
1E+00
1 2
Release height (m)
Figure 163: Effect o f release height and air discharge height on
The above analysis has revealed several important points. Depending upon the conditions V Ft F
varies widely, and small changes in certain variables can produce a dramatic effect. Altering one
single variable that affects C fmax will also influence V Ft F in the same direction but was rarely
directly proportional. Where the conditions are such that Cfmax < LFL, V Ft F “jumps” to
smaller values and was thereon subject to minimal changes.
373
SECTION 4: CONSEQUENCE ANALYSIS
4.1 Selection of consequences

The two selected forms of direct consequence of ignition were thermal intensity ( I ) , which can be
used to quantify degrees of bums to people (or likelihood of secondary fire), and overpressure
( Ap° ) that causes damage to building structure and injury to people. An outcome of high I
and/or Ap° includes damage to buildings, their contents, injury to people and fatalities.
As mentioned in Section 2.1, the outcome is limited to fatality of room occupants, due to either
consequence. A further consequence of ignition is a secondary fire, either caused by direct
impingement of a sustained jet flame on a combustible material, or ignition of materials due to
sufficient / (termed “flashover’). With regards to an ignited jet or plume, its velocity (at C F ) is
generally greater than flame speed and so burning is unlikely to be sustained. The necessary I
for flashover varies considerably with construction and furniture materials. Data collated by
Babrauskas et al (2003) showed flashover to occur over a range of 1 - 125 - 4100 s (kW m 2)4^,
including different foam insulation, wood and gypsum based materials. Other easily combustible
materials such as loose paper or hair may ignite at even lower I . ADL (1998) uses different
probabilities of a secondary fire as a function of the frequency of the prior consequence: from a
jet fire /* x 10'2, from a cloud fire f* x 1O'3, and f * x 10"2 from an explosion. The choice of
these factors was not explained and they seem to be arbitrary'. Therefore, w ithout additional data
to determine the probability of the presence o f various combustible materials, it would not be
meaningful to estimate the likelihood of secondary' fires from flashover. On this basis it was
assumed that the frequency of a secondary fire equates to ignition frequency, for simplicity. The
probability of fatality directly caused by a secondary fire is ignored.
Models to calculate I and Ap° as a result o f igniting flammable material within a room are
detailed below.28 These models demand flammable property' data of refrigerants such heat of
combustion, and these are normally quoted for stoichiometric concentrations. However, the
results of the dispersion model demonstrated that the mean concentration within flammable
volumes were rarely at stoichiometric, generally low er. Flammable properties at concentrations
“8 The overpressure model applies to a flammable mixture in an (empty) room. The case of
ignition of a mixture within a compartment (such as a refrigerated cabinet, or condensing unit
housing) has been ignored for simplicity. However, recent incident investigations (e.g. Blades and
Colboume. 2004) demonstrated that this is a possible outcome, and should be addressed in future
studies.
374
other than stoichiometric differ considerably and therefore for the models described below
properties at a mean concentration corresponding to m F I V F were employed. Data for

flammable properties at stoichiometric concentrations are provided in Appendix 4, along with
methods used to approximate non-stoichiometric properties.
4.2 Thermal radiation model

When a flammable substance bums, the heat energy is partially given off as thermal radiation. A
‘"dose” of thermal radiation may be sufficient to cause bums to individuals nearby. The severity
of bums is a function of heat flux and its duration, and this dose is termed thermal intensity
(equation 387).
Io c c .i Qocc.i ^ flam
where t jjam is the time taken for complete combustion, as calculated below for overpressures
(equation 395). The heat flux radiated onto an occupant ( , in kW m'2) was calculated with a
correlation for a “fireball” from Hymes et al (1996) (equation 388). Other correlations available
for “pool fires” were not considered relevant, given the assumption of vapour-only releases as
discussed previously.
A //(C F ) - < ,
*7 occA ^ (388)
t flam 4 • It • 2\
where the incidence ratio g is influenced by the size and shape o f the flammable volume; g =
0.2 for flames from a plume, and g = 0.3 for fireballs (Hymes et al, 1996). T is the radiative
fraction of the fuel, or the proportion of energy from the burning material that translates to
thermal radiation; the data presented by Beyler (2002), gives T =0.3. A H ( C F) is the HOC at
the corresponding mean concentration based on the average flammable mass over the flammable
time, ( m F I V F ) and t ^ was the time for combustion. z occ is the distance between the fire and
the occupant which is impractical to estimate, but assuming occupants are evenly distributed,
^ = 0.5 / n , where Aocc is the average area occupied per person. Normally, a term for
heat absorption in the atmosphere (by water and carbon dioxide) is included in equation (388),
but since the distances are relatively small, this factor was neglected.
At around I ^ = 115 s (kW m'2)4/3 pain will be felt on exposed skin, becoming more severe with
higher I ^ . Above I occ = 1050 s (kW m"2)4° fatalities are expected and death is almost
375
guaranteed at 6500 s (kW m'2)43. The probability of fatality' associated with a dose of thermal
radiation was calculated using data from Hymes et al (1996), by equation (389), when 0 < Pfatal
< 1.
P fa ^ o c c ) = ^ b - \ n l occ (389)
where a = - 3.79 and b = 0.54.
4.3 Overpressure model

Modelling an explosion from ignition of a flammable mixture is complex, and the most accurate
methods typically employ CFD codes. As with estimation of flammable quantities, use of CFD
was not considered suitable so a more simplistic lumped approach was used.
When a flammable mixture is ignited, the exothermic reaction raises the temperature of the
combustion products, translating to an increase in pressure and/or volume. Under ideal
conditions, where ignition is at the centre of a spherical stoichiometric mixture, the maximum
pressure or volume can be reliably determined from the gas law (Harris, 1983). Transposing the
gas law, a non-dimensional expansion ratio ( E ) is obtained (equation 390).
£ _ NZ'm 273 + 7 V m
flam _______ flam (390)
" Ng* ' 273 + 7;
where Ta is the initial (ambient) temperature of the flammable volume, and Tflam is the maximum
adiabatic flame temperature. and N™ 1 are the number of moles in the burned and
unbumed mixture, respectively. Assuming a confined (fixed) volume that is completely filled with
the mixture, the maximum rise in pressure is the product of initial (atmospheric) pressure ( p atm)
and E . The pressure developed within the space ( Ap ° ) is the difference between the initial and
maximum pressure (equation 391).
AP° = P a m (1 + E ) - P a t m (391)
Conversely, if the flammable mixture were in an unconfined volume (i.e., constant pressure) its
volume after combustion would be equation (392).
V/ , ™=Vf -E (392)
Although for the situation under consideration, the flammable material is w ithin a confined space,
that V F < VRm implies both a pressure rise, and a volumetric change of the burning mixture
376
within the room. As a result the overpressure within the room ( Ap°Rm) will be less than equation
(391). Assuming a fixed room volume, and substituting equation (391) into the gas law, ApRm
was calculated from the ratio of mass of flammable mixture ( m Fa - m F + m F ) and total air
mass plus refrigerant mass within the room ( mRm tot) (equation 393).
E — (393)
^ R m jo t
Harris (1983) and others show that the pressure and/or volume of a flammable mixture increase
exponentially following ignition until a maximum pressure ( p max) is reached, and that
development of Ap° at time ( /) after ignition follows the “cube-law" (equation 394).
K■ -u \nm •t3
Ap°(t) = ---- — ---- (394)
î
where the constant K is dependent upon the characteristics of the explosion and w as found by
rearranging equation (394), and setting Ap° = Ap°niax = ( p max - p am) and t as the combustion
time, tflam. tflam was estimated for the flame travel distance ( 2 llam) from the point of ignition to
the point at which combustion ceases (equation 395).
* flam ~ 2 flam ^ U flam (3 9 5 )
If ignition occurs at the centre of the volume, z flam is the mean distance from the centre to the
boundary of V F , or alternatively if ignited at some distant point, say in the room comer, z flam is
the furthest distance between the boundaries o f V F . The burning velocity { uflam) is based on
laminar flame velocity' (Uflamjam) which is a property' o f the gas, and for HCs it is around 0.4 m s'
1 (Appendix 4). With a rise in pressure or volume, the flame front accelerates and under ideal
conditions, flame speed increases proportionally with E . However, the influence o f room
geometry, obstacles and air movement creates turbulence which further accelerates the flame
front. Therefore, an additional turbulence factor ( T ) was included in the determination o f burning
velocity (equation 396).
Uflam = Ufla m ,la m ' E 'T (396)
Correlations exist for estimating T for “congested” situations (e.g., chemical plant), but a
literature review revealed nothing for relatively empty spaces such as occupied rooms. Instead, a
377
general value of T was obtained by adjustment until optimum fit of model results with published
experimental data was obtained. The average ratio Ap°calc / hp°meas approached unity when T =
2.94, and this also corresponded with the best fit (r2 = 0.83). Conventionally, T is defined as the
ratio of the turbulent flame front area, to that of a laminar (and spherical) flame front.
Equation (393) is valid for a completely sealed enclosure, but realistically, most occupied spaces
have paths in the room fabric through which developed pressure will escape. As &p°Rm rises,
material may be vented from the room, and this occurs in two stages. Initially, any openings such
as gaps in the room fabric, ventilation ducts, etc., allows the higher pressure gas to escape from
the room. Secondly, certain barriers that comprise the room construction may eventually give
way at a given &pRm, thereby enlarging the venting area. In both cases, the exhausted material
results in a lower rate of pressure rise. Using the steady flow equation (397), the mass of material
vented through an opening of a given area ( Av) can be estimated, assuming that the velocity of
the vented material ( uY) is less than sonic velocity (u ^ ).
(397)
where p v is the density of vented gas, which may comprise room air, combustion products and
gas/air mixture. For simplicity, p v was calculated according to the mass fraction of the gas
components at time t . Under certain conditions, the vented gas may approach uss, in which case
equation (398) was used, i.e., u^ = wv.
(398)
The overpressure required to blow out vulnerable barriers within the room fabric w as determined
using a method for explosion vent panels. Hochst and Leuchel (1998) conducted experiments on
the blow-out of doors and windows, and found that such equations for explosion vent panels
could be applied. Both Anthony (1977) and Razus and Krause (2001) reviewed a number of
different approaches. In general, all studies report that the blow-out pressure is related to a
number of specific parameters: fixing force of the vent panel ( p vPtfix), plane density of the panel
( Wvp), burning velocity and a ratio o f wall to vent area ( A wall / A^ ). Razus and Krause found
that the correlation of Molkov provided the best results compared with experiments. An improved
approach was developed by Molkov et al (2003, 2004) that suggested equation (399).
378
A wall Wvp
flam 1/3
(399)
"vp V,
r Rm
where the coefficient a =0.00043. m is a “deflagration outflow interaction'’ parameter, from

equation (400) for HC-air mixtures, representing the inertia of the panel.
0.4
0 +10 - v ^ ) - (1 + 0.5-JO ?)
E7 = 1.75 - (400)
l + Pvp,Jix/ P atm
where Br is the Bradley number (equation 401).
Ap
Br = (401)
VRm
2,3 ■u flam.lam ( E - l )
Thus, once ApRm > &P°-P^ the area Avp becomes available for venting material. However, as gas
exits the room &p°Rm is affected, and this is influenced by the quantity of high temperature
combustion products that no longer contribute to a pressure rise. So, throughout the combustion
process the change in ApRm was based on the total mass of material present in the room, and the
corresponding change in average room temperature ( TRm) (equation 402).
ApL(D = Apim(0) (402)

\ ^ R m ,to t (0). [273+ 7 ^ (0 )]
where (0) indicates initial conditions, and TRm is based on the mass of each mixture, and
assumed similar specific heats (equation 403).
_ K a ( 0 W 2 + TrFa] + mflaJ t ) -[273+ Tflam] + mrJ t ) -[273+ TrJ

TRm(t) = ------------------------------ 1-------------------- 273 (403)
m Rm,tot
where mRmtot is the sum mass of unbumed flammable mixture ( mâ ), burned mixture ( mflam)
and non flammable refrigerant/air mixture ( mr a ), whilst accounting for the proportion having
escaped through vents (equations 404, 405 and 406). The mass of mixture burned by time t , is
the product of mixture density and volume obtained by the transposition of equation (394).
/* \ P flam ’ K ' ( P flam ’ 0 v .

mflam(t) = ------------— -------------X flam{t )-mv{t) (404)
379
^ F /rv\ P flam ' ^ ' ( U flam ' 0 ^
wr,a ( 0 ) - --------- —--------- (405)
™ r,a ( 0 = " V * ( 0 ) - a (0 •Wv(0 (406)
where X is the mass fraction of the given mixture within the room at time t . As unbumed
flammable material is vented, the explosion ceases before the anticipated burning time (as
equation 395), and therefore a revised tflam was determined as /wrF0(f) —»■0. This would be
relevant in the calculations of l occ, reducing the intensity.
1E+03 § 600% T
□ DeHaan
Cu o o Razus #1
$ 400%
a Razus #2
Os 1E+02 -
o Razus #3
%
200%
1 o DeHaan
JS 1E+01 - %
a Razus #1 i _o A.o- - ô-o-Q- %-
& 0% oôoo ocr
-
o Razus #2
a Razus #3
1E+00 V? -200%
1E+00 1E+01 1E+02 1E+03 1E+00 1E-+01 1E+02 1E-K)3
Measured Ap0^ (kPa) Measured Ap ^
0 (kPa)
Figure 164: Comparison o f measured and Figure 165: Deviation between measured and
calculated maximum overpressures calculated maximum overpressures
As mentioned earlier, the model was tested against experimental data for vented explosions as
cited in Razus and Krause (2001) for methane/air and propane/air mixtures, as w ell as small
hexane clouds from DeHaan et al (2001). Hjertager et al (1998) also reported data for vented
explosions from stratified layers, but no information on flammable quantities or distribution of
concentrations was provided so it was not possible to utilise the information. Both measured and
calculated t±p°Rm 1S shown in Figure 164, and all relevant data for the measurements are provided
in Appendix 5. The different data-sets from Razus and Krause correspond to the different gases
and room sizes by which they compiled the measurements.
Following the adjustment of T (to 2.94), the model predicted maximum tsp°Rm to within ± 100%
for 40 of the 45 measurements, and within ± 50% for 31 of the measurements, wfiich were
380
considered good agreements. As seen in Figure 164, three of the conditions caused the model to
over predict by 4 - 6 times, which was inconsistent with the remaining data. It was concluded
that this relatively simple model provided a sufficiently accurate estimation of overpressures for
the purposes of this QRA.
Four examples of &p°Rm development over time are given in Figure 166, which were based on a
flammable R290 mixture in a 40 m2 * 3 m room. By far the greatest ApRm is for a fully confined
room that is entirely filled (120 m3) with a stoichiometric mixture. By introducing a vent panel
(equivalent to 5% or the total wall area), there is a major reduction in &pRm as the excess
pressure is rapidly exhausted, noting also that the bum time is reduced because some of the
flammable material was vented before it was burned. A more realistic situation is where a 24 mJ
mixture (at a concentration between LFL and stoichiometric) was ignited, and produced &pRm
proportionally lower than the completely flammable room. Again, by introducing a vent panel
(refer to right-hand scale), Ep°Rm is further reduced; this example shows a temporary' reduction in
ApRm(t ) at the point where the vent gives way, shortly after ApRm(t) began to rise again.
-~r 20
0 100 200 300 400

Bum time (ms)
Figure 166: Examples o f room overpressure development
To interpret the severity, any Ap°m below about 3 kPa is unlikely to cause any physical damage
to persons or property. Persons exposed to overpressure (alone) above about 250 kPa would be
begin to suffer fatalities. As &pRm approaches 450 kPa, death is almost guaranteed. Probability
o f fatality due to overpressures was calculated using data from Beyler (2001) in equation (407),
381
Pfa*a(*P°Rm)=a + b -fobP°Rm (407)
where a = - 3.62 and b = 0.76.
4.4 Application
Some of the results from the dispersion model were used to generate I ^ and ApRm, had the
flammable mixture been ignited. Certain assumptions were made: z flafn was based on a central
ignition point, an occupancy of 1 per 10 m2 of floor area, and initially the room fabric was
resistant to overpressure (i.e., no venting). The results are shown in Figure 169 to Figure 168,
where I ^ was indicated with solid data-points and Ep°Rm with hollow data-points.
s 1E+04 f 1E+04
i
~ 1E+02 - 1E+02 -
5 1E+00 1E+00
c£ 40 m2 mr = 1.0 kg
.j 0.008 kg m-3 C. = 0.008 kg m-3

IE-02 > IE-02
0.5 1 1.5 2 2.5 0 20 40 60 80 100
Released mass (kg) R oan area (m2)
Figure 167: Effect o f release mass with Figure 168: Effect o f room area with constant
constant room size or constant CRm on release mass or constant CRm on consequence
consequence
Figure 167 and Figure 168 give the results for the effect of released mass (for constant room area
and fixed CRm), and room area (for constant mass and fixed CRm), respectively. In Figure 167 it
was seen that larger mr alw ays increased the severity o f the consequence, although for fixed
CRm, raising mr above about 1 kg produced little further change in I occ and Ep°Rm. Compared
the resulting change in V Ft F (Figure 159) the sensitivity of I occ and ApRm on room area seem
to be less pronounced. This was attributed to the fact that whilst V Ft F increases, it is only due to
an extended t F whereas the “instantaneous” m F (which dictates I occ and A p L ) becomes
382
almost constant. The same tendency of minimal change in I ^ and Ep°Rm at larger room areas
was seen in Figure 168 where room size is varied (compared to the change in V Ft F , Figure 160).
The effect of varying room airspeed is shown in Figure 169, where air discharge angles were
upwards and downwards - the data for horizontal airflow was not included for clarity. In general
it was seen that the results for both I ^ and &pRm follow a similar pattern, which also follows
V FtF when compared with the results in Section 3.4.8. However, I occ appeared to be slightly
more sensitive to the variation in uRm than ApRm. Similarly, a downward directed air discharge
(0 = n I 3) significantly reduces the severity of both consequences. Figure 170 provides similar
data for different release height (where hd was constant at 2 m). Elevating the position of hr
always resulted in a reduction of 1^ and &pRm. Although not included here, the results for hd =
hr also followed the trend found for V Ft F .
1E+04 T 1E-K)4 -r—
'aT 1E+02 ' IE-02 -
Rm
o
1E+00 - IE -00 -
uRm = 9.03 m s 1
= 0.08 m s - 1
^ IE-02 IE-02
0 0.05 0.1 0.15 1 2 3

Release hek jt (m)
(m s - )
Figure 169: Effect o f room airspeed on Figure 170: Effect o f release height and
consequence airspeed on consequence
Regarding thermal intensity, the most relevant independent factor is the distance (as equation
388), which was considered to be a function of population density. Under any set o f conditions, a
more populated environment produced a rise in I occ proportional to the mean target distance to
the power % (as equation 387).
383
Throughout, the severity of the consequence was mostly proportional to V Ft F due to the
dependency on m F . When explosions develop high overpressures, the values indicated above are
unlikely to occur due to the venting phenomenon. For example, most occupied rooms are
constructed with window's, doors and gaps in the fabric that could act to vent an explosion. To
investigate the differences in consequence associated with venting, the conditions used to provide
the results in Figure 169 to Figure 168 were used to recalculate Ep°Rm when a number of vent
situations were employed. Results are given in Figure 171 as the ratio of vented to unvented (or
confined) ApRm for four different venting situations, using m F as the reference variable. The
different forms of venting were the presence of gaps in the fabric that normally lead to
infiltration, an open door, and windows that occupy 3% and 8% of the rooms’ total w all area.
to blow-out w in d o w ____
o open door
□ 3% window
a 8° o window
0.4 - □o
c
£
£
0.001 0.01 0.1 1 10
n>r (kg)
Figure 171: Effect o f various venting routes on reduction o f overpressure
With gaps corresponding to an infiltration rate of A50 = 14 h 1 (equivalent to a total vent area of
about 0.1 m2 in a 3 m x 40 m2 room), Ep°Rm was reduced by about 10% of that of the confined
space. An open 2 m2 door, reduced ApRm to approximately 10 - 40% of unvented Ep°Rm. but as
m F exceeded about 0.1 kg, the vented JSp°Rm reduced to below' 5%. In the case of window's,
using typical values of p ^ fix =4.5 kPa and = 7.5 kg m 2, the variation of vented ApRm wras
different from the open (door, infiltration) vents. An initial resistance is offered by the window
fixing, so that regardless of the severity of the explosion, the presence of any sized window would
not permit reduction in overpressure below that 4.5 kPa (corresponding to m F ~ 0.05 kg, as
indicated in Figure 171). Above this value, plus a minor addition associated with the inertia of the
glass (panel), a considerable reduction in ApRm occurred. In the case of a large w indow area (8%
384
of the total wall surface), an exponential fall in ApRm was seen, until it approached a few percent
of the confined &p°Rm at the largest m F . Where a smaller window was present (3% of the total
wall surface), a similar reduction occurred, but varied by up to 60% of the original value. This is
due to the vent area being insufficient to relieve the majority of the developed pressure under
certain circumstances.
Comparing the minimum I ^ and hpRm to achieve Pfatal > 0, against the results in Figure 169
to Figure 168, showed that fewr conditions produced a consequence severe enough to cause
fatality. Only 0.5% of the cases present a probability’ of fatality greater than zero, and these were
due to thermal intensity only. If the occupancy density’ in the calculations were doubled (to 1
person per 5 m2), then the number of situations for potential fatality rose to 8%. Were the severity
of injury' due to I ^ considered, around 10% of the cases could lead to second-degree bums (>
700 s (kW m'2)4/3, Hymes et al, 1996). None of the calculated &pRni - with or without a means of
venting - would lead to death. Were injury from &pRmconsidered (Beyler, 2001), lung damage (>
85 kPa) could potentially occur in about 5% of the (unvented explosion) cases, and eardrum
damage (>35 kPa) in around 30% of cases. Data is also available to evaluate the damage to
parts of buildings (e.g., Harris, 1983).
385
SECTION 5: APPLICATION OF OVERALL MODEL
5.1 Uses for the QRA model

The individual components of the entire QRA model were detailed in previous sections. Each sub
model was linked into a single program module, and compiled in a sequence according to Figure
172. A single set of inputs were required, and shared amongst relevant calculations. Outputs were
generated for several purposes: evolution of the overall risk showing the influence of each
component, parametric assessment of key input variables, example calculations for selected RAC
equipment, and lastly derivation of a simplified risk calculation method.
Evolution of the risk calculation is illustrated with output data from each probability or frequency
sub-model, and the intermediate stages where these are combined. The parametric evaluation
showed how' the risk is affected by changing parameters associated with the equipment design, the
installation environment and aspects associated with its operation. Example cases are given for
four different equipment types, considering the variation of risk when equipment is designed
according to a safety standard.
Prior to these exercises, the issue of validating the QRA model was considered. In general, it is
virtually impossible to empirically validate a QRA since not only is comprehensive historical data
needed (which should include detailed records of equipment design, installation conditions and
incidents), but there must be a large number of installations of identical design, operating and
environmental conditions. Unfortunately, with an emerging technology (such as HC refrigerants),
such data is scarce. Validation becomes more difficult when considering that the majority of
consequences are relatively minor (e.g., small jet/flash fire) and of no concern to the relevant
authorities, making reporting unlikely. Nevertheless, information about all known incidents has
been collated for consideration (Table 49). Initial observations from Table 49 were that most
accidents occurred either during servicing or as a direct result of the servicing activity. Also,
about half of the incidents are associated with equipment that was retrofitted with HCs, and
therefore not designed appropriately. Nevertheless, in absence of market data for most of the
equipment types above, it was not possible to compare the incidence of reported ignition events
for a known population, against risk frequencies from the model.
386
A.
activity:
normal operation'storage/servicing
infiltration model weather/ building data
for leak size = 1 to N

for operating mode = on/off
for infiltration rate = I t o N
for thermal convection = on/off
for unit airflow = o n o ff
for mechanical vent = on/off
thermal source, unit and

airflow models mechanical airflow data
installation/leakage data
plume/cloud flammable volume and

time models
thermal radiation
overpressure model
model
time fraction and probability model ignition source data
ignition and fatality' frequencies
___________ T
( End )
Figure 172: Overall modelling sequence fo r the combined ORA
Another point of interest is comparison of results produced by the present QRA, against the
results of QRA exercises reported in the literature. With reference to the studies discussed in
Section 1.2.1, it was decided that a direct comparison was not possible, either because the
application was outside the present scope, or because insufficient input details were provided for
a precise representation of the equipment and installation characteristics. The latter was
confirmed by applying a range of assumed input values to the QRA model, to find that it output
frequencies that varied over several orders of magnitude for the single application. However, in
conducting this exercise it was found that the range of modelled risk frequencies for a given
application was typically lower than the ignition/fire and fatality- frequencies of the published
QRAs. This observation implies that the more general QRA methods previously employed
387
elsewhere overestimate the risk, which was attributed primarily to the manner by which
assumptions are often made within the context of a safety assessment. Normally, an assumed
quantity represents an educated guess at an “average"’ value in absence of better data, but with
safety assessments the conventional approach is to skew assumptions towards values that would
lead to a more severe outcome. Thus, where existing QRAs adopt assumptions for inputs or
calculation methods, they tend to employ “worse case” values. Since the current study sought
empirical data and correlations to overcome many of the assumptions used within published
QRAs, the choice of exaggerated values for variables that could increase the risk was avoided.
Whilst the present study does rely on a limited number of least favourable suppositions (i.e.,
vapour-onlv, low' velocity leak), the series of worse-case assumptions adopted at each stage in the
calculation sequence (in most of the reported QRAs) compounds the overestimation. It is the
avoidance of this effect that makes the risk estimates from the present QRA significantly lower
than the literature.
Table 49: Reports o f incidents involving HC refrigerants
Description Equipment Cause SOI Effect Reference

Home. Domestic Leaked from joint Thermostat Fire and Workplace
Australia fridge. that was made overpressure Services.
retrofitted with epoxy resin. sufficient to 2000
distort fridge
Cafe. Commercial Overcharged Spark from Flash fire, Workplace
Australia fridge, system known to plug when bum injuries Services.
retrofitted have a leak, pulled from 2000
condenser split socket
when charging.
Cellar, Domestic Leak from Service Temporary jet Afjei, 1999
Switzerland heat pump evaporator into person’s flame
water circuit and cigarette
exited via air vent lighter, used
when looking
for leak
Workshop, Domestic Broke into system Brazing torch Fire and Mariager,
Algeria fridge unaware of minor 2001
flammable damage
refrigerant
Workshop, Mobile air Release during Not known Small flash Private
Philippines conditioner, servicing (possibly fire/jet flame communicati
retrofitted lighting on, 2002
cigarette)
Home, Domestic Leaked into Thermostat Overpressure WMFS, 2003
England (3 fridge refrigerated sufficient to
cases) compartment distort fridge
388
5.2 Evolution of risk calculation
The probability models in Section 2 were employed to illustrate the results of the component
probability’ calculations, and the evolving overall risk frequencies. To do this a set of reference
conditions was chosen and the results from the probability and frequency calculations (based on
results produced with the dispersion model in Section 3 and consequence models in Section 4)
were presented. Calculations were based on an installation according to the following: 1 kg of
R290 in a system comprising 30 joints, where hr and hd are both at 2 m. Following a leak, none
of the charge was retained within the system. It operated for 12 hours per day, and during on-
cvcle, Vd = 0.033 m3 s'1 (the low’ value - equivalent to A = 1 h'1 - was chosen for illustrative
purposes); the forced airflow7remained constant in the event of a leak, and mechanical equipment
faults were ignored. The room was 40 m2 * 3 m high and infiltration = 14 h'1, where the
corresponding A infiki were as Figure 101. There were four occupants present for 24 hours per
day, although airflow from thermal sources was neglected. Within the lower (0 m - 1 m) and
middle (1 m - 2 m) room levels, there w?ere five SOI and two SOI, respectively, and all SOI were
active for 1 s, four times each day in a random manner. Other probabilities associated with the
release were: PIeak = 1, P ^ = 0.73, Pmiajl = 1 and P^s = 1. For the consequence, the room was
totally confined (except for the infiltration opening of 0.084 m2), and ignition always occurred at
the centre of the flammable volume.
Following the sequence in Figure 172, leak frequencies and leak times were characterised,
corresponding flammable quantities were obtained from the dispersion model, and then the
component probabilities - P Fm, P F and P F* - w ere calculated, follow ed by the severity of
consequences and associated xIocc, %Apo and X . Information on system operation, occupants and
weather conditions w7ere used to establish the relevant time fractions.
f leak and tleak were calculated for 15 different sizes (as Section 2.3) for both on- and off-cycle.
For each leak, V F , m F and t F were generated for both on-cycle (when Vd = 0.033 m3 s'1) and
off-cycle (Vd= 0), and for each cycling mode the calculation were carried out for (additional)
airflow? from infiltration for the range of A mfil. Subsequently, P F

01, P F , P F*and finally f *
were obtained for each separate combination of conditions: cycling mode, leak size, forced
airflow, and infiltration rate. Results for f leak, P F

o i, and P F at the lowest infiltration rate ( A infll
= 0.4 h 1, chosen to clarify variations) are presented in Figure 173, for both on- and off-cycle over
389
a range of leak times, or leak sizes (on the basis that a shorter leak of the same mass produces a
‘‘larger” leak); the data does not include adjustment with .
Generally for longer leak times ( tleak > 2000 s), f leak for a given leak size was similar for both
on- and off-cycle. The observed deviation at short leak times was expected given that their
duration would normally be shorter than compressor cycling intervals for many equipment types.
The trend in Pfm behaves in a similar fashion. For longer leak times, the similarity between on-
and off-cycle is due to the feet that the released is dispersed within tleak, so t F = tleak . This is
consistent for all leak sizes during on-cycle since the associated forced airflow is sufficient to
rapidly disperse the release even for the most catastrophic leak. Conversely, for tleak < 1000 s,
the conditions during off-cycle cause a flammable cloud such that t F > tleak, and P f tends to
increase towards a catastrophic release.
1E+00
(on-cycle
hollow,
off-cycle •
filled)
IE -12 -I------------------- 1--------------:-----i

1E+02 1E+03 1E-+04 1E+05
Leak time (s)
Figure 173: f eak, PFsoi, PFv and f * fo r different leak sizes at Ainfu = 0.4 h'1
Another factor that affects P f is the positioning of the SOI - as tleak increases, the region
occupied by V F gets smaller and gradually shifts closer to the origin of the release, i.e., upwards
towards hr . Consequently, the possibility of ignition becomes more dependent on the SOI within
the middle level. This is seen in Figure 174, which plots the ratio of local ignition frequency in the
lower and middle vertical room elements to the total frequency ( / fc* / / * ) for the different leak
sizes (the upper element is ignored since C F never occurs there) during off-cycle only. It is seen
at shorter leak times, the ignition frequency is mainly attributed to the lower level, and this is
because the corresponding large mass flow rate creates a stratified layer on the floor, thus large
390
P f and Py . As tkak gets larger, the greater proportion of the ignition frequency occurs at the
middle level, and the contribution at the lower level falls off. The distribution is equal at about
tleak ~ 5* 103 s, and the entire / * shifts to the middle level at about tleak ~ 104 s, when C F cease
to exist in the lower level (so for the lower level, P Ff = 0, P f = 0). A similar trend in the
distribution of / * was seen for on-cycle, although the transition from / * dominating the low
level to the middle level occurred at shorter leak times since the higher airflow disperses larger
releases more rapidly before they reach floor level. These observations demonstrate the
importance of establishing the correct position of the SOI and V F . Also shown in Figure 174 is
the ratio of total / * to the ‘'lumped'' ignition frequency which was calculated assuming a single
element, i.e., Pfm and PF were calculated for the total V FtF . The results further support the
importance of discretising the room volume, as in the present case the lumped approach
overestimates f * for a catastrophic leak by 40% and underestimates f * for the longest leak
time by 90%.
1E+01 i IE-06
= 325 s
IE-01 -
= 777 s
-- 1&10 -
IE-03 -
-¥r- lumped (on-cycle - hollow,

off-cycle - filled)
IE-05 1E>14
1E+02 1E+03 1E+04 1E+05 1 2 3
Leak time (s) Infiltration rate ( h 1)
Figure 174: Variation offh* with room level Figure 175: f* fo r different leak times for a
fo r different leak sizes fo r o ff-cycle range o f infiltration rates
The overall variation in / * with tleak is also shown in Figure 173. For longer leak sizes, it is
seen that / * reduces slightly from about tleak ~ 103 s and longer, and this applies to both on- and
off-cycle whilst having similar values. This trend is expected when observing the component
probabilities, f leak, PfOI, and P f , for each leak size. For shorter, more catastrophic leaks, a
deviation between on- and off-cycle / * occurs, although under both conditions, individual f*
391
are significantly greater than when tkak > 103 s. These higher / * follow the higher values of the
component P F, , and P F which are due to the larger V F resulting from the catastrophic leak
conditions.
Next, the effect of different infiltration rates was considered. Figure 175 shows how f * (not
adjusted for <f>wjU, ) varies with A ltrfil, for a number of different leak sizes up to 800 s (since the
trend thereafter is obvious). At low’ h infil and short leak times, high / * occur and reduce rapidly
with greater A infiI. Eventually, further increase in A infil has negligible impact on f * as it remains
approximately constant thereafter. The infiltration rate at which this transition to constant / *
gets lower as leak time is prolonged. This is explained by reference to Figure 161 where once a
particular uRm has been exceeded, V FtF thereon remains nearly constant and Figure 162 where
VFt F also remains constant for longer leak times. Since P F

OI, and P F are strongly dependent
upon V Ft F and t F, it follows that the corresponding / * will follow7this behaviour. In a
practical sense, where high infiltration rates are normally present, evaluation of a single long leak
time for a range of high A mfil may be acceptable. This approach is supported for the present
example because higher A infil occur less often (Figure 101). Lastly, a notable difference is seen
in / * between on- and off-cycle operation, where a constant / * occur at lower A mfil (for the
same leak time) with off-cycle on account of the associated Vd giving higher uRm .
From the data-sets as above, I occ and ApRm w ere calculated as shown in Figure 176. Both I occ
and &pRm are seen to reduce with longer leak time due to smaller V F and m F and likewise,
values for releases during on-cycle are smaller than for off-cycle. A transition in the gradient
occurs when there is no longer flammable cloud development at the floor and thereon I occ and
ApRm are similar regardless of operating mode. Under conditions that lead to highest 100
s (kW m'2)4'3), ignition may cause minor bums and may be sufficient to produce secondary
ignition of easily combustible materials. The highest ApRm would damage windows and light
panels but is insufficient to injure occupants. Ignition of a release longer than 500 s would
produce I ^ that is unlikely to be felt by occupants, and ApRm w ould not damage occupants or
property. Under no circumstances would the consequences lead to fatality. (The values below7 10"4
392
s (kW m'2)43 were ignored because of negligible practical significance - for example, I ^ = 1CT4
s (kW m'2)43 would be comparable to holding a cigarette lighter at arms length for one second.)
1E+02 7 ■T IE-05
I occ (on-cycle) I x c r (on-cycle)
I occ (off-cycfe) - I c c c ' f (off-cycle)
ApL (on-cycle) 'r IE-09 ^P°Rm - / ‘ (on-cycle)
Ap L. (°ff 1cycfe> ApL / * (off cycle)

t IE-01
IE-13
IE-04 * IE-17
1E+02 1E+03 1E+04 1E+05 1E+02 1E+03 1E+04 1E+05
Leak time (s) Leak time (s)
Figure 176: Overpressure and thermal Figure 177: Integrated consequence-

intensity fo r different leak times frequencies fo r different leak times
Considering the wide variation of I ^ and t±p°Rm due to ignition of small and large leaks, the
value of f * does not necessarily reflect the “risk” because it ignores the severity of the
consequence of the ignition event. Thus, by employing the product / * x I ^ and / * x &pRm as
X locc and Xhpo respectively, the risk is better represented since it accounts for both likelihood of
ignition and the subsequent severity of the consequence. Figure 177 shows these values for the
current example. As with the variation of / * with tleak, the catastrophic leaks produce a notably
higher risk than the longer leaks, although the range over which X io c c Z & po differ is far
greater than f * . The range over which f * varies is in the order 103, whilst for Xiocc and X&po
the variation is about 1010; whilst being subjective, this is believed to provide a better
representation of the difference in risk between an occasional temporary flame, and a more likely
explosion event capable of secondary fire, building damage and serious injury.
For the overall ignition frequencies, individual / * was adjusted for <f>wfiu, summed for all
infiltration rates, and these were adjusted to account for 50% on-cycle (<t>op = 0.5) and 50% off-
cycle time fractions. The resulting / * for on-cycle, off-cycle and the summed total / * is
presented in Figure 178 for the range of leak sizes. Compared to Figure 173 and Figure 175 for
393
A infil = 0.4 h"\ the variation in total / * is seen to be smoother, where there is less o f a
distinction over the range of leak sizes, due to the addition of frequencies for the higher
infiltration rates, since these are known to have more consistent / * as seen in Figure 175.
However, despite the evening effect of the contribution of higher infiltration rates to the overall
/ * , the previously observed elevated / * associated with catastrophic leaks during off-cycle still
dominates.
IE-06 -i IE-06 T
IE-10 - * IE-10 -
^ IE-14 - IE-14 -
on-cycle
oiLcjcle
total
IE-18 IE-18
1E+03 1E+04 1E+05 1E+02 1E+04 1E+05

Leak time (s) Leak time (s)
Figure 178: Overall ignition frequencies Figure 179: Overall risk factor
The overall risk factor as a function of leak size is shown in Figure 179, again for off-cycle, on-
cycle and the total X (handled the same as Figure 178). This data follow s the previously
identified characteristics of Xiocc and Z apo Figure 177, where on- and off-cycle X are similar
with long tleak, and that the off-cycle X dominates the total for catastrophic releases. Most
importantly, the variation of X with leak size compared to / * (in Figure 178), implies a
relatively greater risk for catastrophic leaks than for “pinhole” leaks.
For the installation as a whole, on-cycle / * = 3.2x1 O'9, off-cycle / * = 4.9x1 O'9, and the (sum)
total / * = 8.1 x 10"9. The comparable values for on- and off-cycle imply that the risk (per
operating time, since <pop = 0.5) is similar regardless of operating mode. However, in terms of
risk factor, on-cycle X = 3.4* 10'9, off-cycle X = 3.6* 10'8, and total X = 3.9><10'8. Thus using
X for comparative purposes shows that that the risk during off-cycle accounts for about 90% of
the total risk, and therefore represents a more hazardous situation per unit of time. Thus X is
394
considered to better characterise the extent of the risk associated with the range o f conditions
encountered.
Table 50: Comparison with maximum acceptable risk criteria
Criteria Example results Limits (members of public)
Frequency of ignition (fire) ( / * ) s .ix io V 1 < lxlO ’V
Thermal intensity ( I ^ ) 79 s (kW m'2)4/3 < 1050 s (kW in 2)43
Overpressure ( &p°Rm) 15.4 kPa < 250 kPa
Frequency of fatality ( f fatal) O y1 < lx K fV 1
Integrated thermal intensity-ffeq. ( j /occ) 1.6 *10'7 s (kW m'2)4'3 y’1 < 5.6*1 O'3 s (kW m'2)43 y'1
Integrated overpressure-ffeq. ( x Apo) 3.7* 10'8 kPay'1 < 4.2* 104 kPa y 1
Risk factor ( X ) 3.9*10'8 < 5.9*10'4
Finally, the results were compared against the various criteria for determining the significance of
the calculated risk measures (Table 50) . In terms of / * , %]occ, x^po ar*d X , calculated values
are in the order of 103 to 104 times lower than the maximum tolerable values for members of the
public (from Table 36). f fata] is zero and I ^ and &p°Rm are around 10% of the maximum
values.
5.3 Param etric assessment of input variables

A parametric evaluation was conducted to indicate how certain variables would affect risk, based
on a fixed set of input (reference) values for a split air conditioning system in an office
environment, similar to the situation detailed in Section 5 .2. Values associated with the equipment
design, operation and the installation environment were addressed, and relevant inputs listed in
Table 51.
Several inputs were not subject to change, either because their influence on the overall risk is
known to be negligible or approximately proportional. Additionally, it was assumed that there
was no synchronisation between occupancy, cycling, SOI events or thermal sources. The fixed
inputs were as follows: location of components as split type system, Pleak =0.21; system leak
probability, = 0.2; unit airflow discharge area, 0.12 m2; number of SOI on unit = 10; unit
SOI active time is 1 s; number of events of each unit SOI = 10 per 24 h; number of SOI = 6
395
(low). 4 (mid), 2 (high); unit airflow fault probability = 0.10; unit airflow fault repair time = 7
days; height of thermal sources is 1.5 m; mechanical ventilation fault probability = 0.10;
mechanical ventilation fault repair time is 30 days: termination of unit operation (compressor and
forced airflow) occurs when 50% of charge is lost.
Table 51: Reference values and variations
Parameter Reference Additional range tested

Refrigerant fr^pe R290 R600a, R1270
eo Refrigerant mass 1 kg 0.5 kg. 1.5 kg
3s
*«3 Number refrigerant circuits 1 2
1 Release height 2m 1m
Te3
re Airflow rate 1h 1 2 h'1
P
<D
P Airflow discharge height 2m 1m
cl c
•p o
O' Airflow discharge direction Horizontal Downward, upward
<L> 5C
i P
o .*3 No of connections 30 10
C O
o CpD
>
• -
CO Unit SOI P ^ , 0.14 1
•C
o Room size 3 m x 40 m2 3 m x 20 in2, 3 m x 60 m2
2
Jl
o Mech vent (airflow/discharge height) 0 h‘V - 3 h'V2.5 m. 3 h'V0.5 m
"ore
’co Mech vent detector initiation - 20% of LFL (as EN 378)
K Window size 4 m2 Om2

Fabric leakage 14 h 1 7 h'1, 28 h’1
No. SOI events/24 h (low. mid. high) 2, 4 .4 4, 8 ,8
CO SOI active time 1s 100 s
re
S to
.2 ‘C Post-ventilation period 0s 300 s
2 .
&2 Unit airflow initiated by reft, detection 20% of LFL (0.008 kg m’3)
o Unit airflow initiated by refrigerant loss - 25% of charge lost
Number of thermal sources 0 20
Table 52 presents results in terms of total ignition frequency, maximum thermal intensity,
maximum overpressure and overall risk factor; integrated thermal intensity-ffequency and
integrated overpressure-frequency were not included as they are encompassed within X .
(Maximum 1 occ and maximum Ap^m refer to the highest value calculated over all conditions
within a single QRA run.) No results for fatality frequency were presented as f fatal = 0 for all of
the conditions. The difference in / * and X for the different variables in the range listed in Table
51 compared to the reference case, were expressed as logarithmic change, i.e., InA f * =
396
1- In / * / In f * {ref} and InA X = 1 - ln X /ln X{ r e f } , in percentages. This presentation assists
interpretation since many of the variations were several orders of magnitude in difference.
Differences in values of 1 ^ and Ap°Rm were expressed as percentages in the usual way, i.e.,
A^ = 1 - A** / /«* W }, and A = 1 - Ap°Rm / Ap°Rm{ref }.
Several general observations were made from Table 52. First, the behaviour of the variation in
/ * is always followed by X , although it is rarely proportional. Under most situations, a large
change in f * and X was consistent with a large change in maximum I ^ and &p°Rm. Further,
variations in maximum I occ and &pRm were consistent with each other (and with maximum
Cf max), and the direction was always followed by / * or X except when the conditions meant
that maximum I ^ and Ap°Rm remained constant. Overall, there were never contradictions
amongst these risk measures, for example, a reduction in f * and increase of X did not occur.
Table 52: Results o f parametric evaluation
InA
Variation for range in
Table 51
r r 1 occ •> Rm
^ o c c ^ R n , X InA
X
(skWm* (-)
(y'1) 2,kPa) (% ) (% )
(%>
Reference 6.9xlO‘10 0.0 1.1, 3.2 0 ,0 7.0xl0'10 0.0
Refrigerant, R600a 5.1xl0'10 -1.4 1,3.1 -8,-4 5.0xl0'10 -1.6
Refrigerant, R1270 1 .8 x l0 10 -6.3 0.4, 1.1 -63, -67 7.4x10" -10.7
mr = 0.75 kg 7.6xl0'n -10.4 0, 0.1 -96, -98 1 .5 x l0 12 -29.3
mr = 1.75 kg 2.3 xlO'9 5.7 3.3, 13.2 198, 313 7.2 xlO'9 11.0
No. circuits = 2 5.0 xlO'11 -12.4 0, 0.1 -96, -98 2.1 xlO 12 -27.6
hr - 1 m 1.8xl0'9 4.5 2.6, 11.3 138, 255 4.1 xlO'9 8.4
hd and hr = 1m 2.8xl0'9 6.7 2.6, 12.7 138, 298 6.4 xlO 9 10.5
Vd , A d = 2 h'1 6.0xl0'10 -0.7 1.1, 3.2 0 ,0 6.6 xlO'10 -0.3
Airflow 6 d = 2 rad 8.4 xlO 10 0.9 1.1, 3.2 0 ,0 9.1 xlO'10 1.2
Airflow7 9d = 1 rad 6 .0 x l0 10 -0.7 1.1, 3.2 0 ,0 6.6xlO'10 -0.3
U n i t C detect =20% LFL 2.9xlO'10 -4.1 1.1, 3.2 0 ,0 1.2xlO'10 -8.5
Post-vent td = 300 s 6.9xlO‘10 0.0 1.1, 3.2 0,0 7.0xl0'10 0.0
^conn = 10 2.9xlO'10 -4.1 1.1, 3.2 0,0 2.3 xlO'10 -5.3
397
InA
Variation for range in r r h c c ^ lm
A l occ>AP°Rm X InA
X
Table 51 (skWnf
( y 1) (%) (-) (%)
(%) IkPa)
Unit Pâ = 1 6 .9 x l0 10 0.0 1.1. 3.2 0 .0 7.0xl0‘10 0.0
Unit P ^ , = 1, hd = 1
9.9xl0'9 12.6 1.1. 3.2 138, 298 9.1 xlO-8 16.4
m
ARm = 20 nr 4.4xlO'10 8.8 6.6, 37.6 500, 1078 3.0X10-8 17.8
ARm = 60 nr 2 .8 x l0 12 -26.1 0 .0 -97, -99 7 .5xl044 -43.4
A mech=3h-1,Amecfc=2.5m 3 .0 x l0 n -14.8 1.1, 3.2 0 ,0 6.8 xlO'12 -22.0
&mech=3h'1, = 0 . 5 m 2.3xl0*n -16.0 1.1, 3.2 0 ,0 6 .5 x l0 12 -22.2
Window Av = 0 m2 6.9xlO'10 0.0 1.1. 3.2 0. 0 7.0xlO'10 0.0
A so = 7 h'1 1.4xl0'9 3.4 2.5. 9 126, 182 3.2xl0'9 7.2
<N
o
A so = 28 h'1 2.4X10'11 -15.9 0, 0.1 -96, -97 -30.5
X
Ng, = 4. 8. 8 9.8xlO'10 1.7 1.1, 3.2 0 ,0 2.8xl0'9 1.7
tsoi = 10° s 1.4xl0‘9 3.3 1.1, 3.2 0 ,0 3.9xl0'9 3.3
25% loss of mr starts

6 .0 x l0 ‘10 -0.7 1.1. 3.2 0 .0 5.2X10’10 -1.4
vd
K therm = 2 0 3.6xlO'10 -3.1 1.1, 3.2 0 ,0 3.5X10'10 -3.3
Regarding equipment design, refrigerant has a distinct influence where R600a produces a slight
reduction in risk (despite a higher C fmax) but a more notable improvement occurs with R1270.
Both refrigerants have a higher LFL than R290 which accounts for a portion of the lower risk,
and R1270 has a lower density than both and therefore easier to disperse. However, a more
thorough comparison of different refrigerants should account for the variation in system charge
given equal refrigerating capacity. This is particularly important with respect to the sensitivity to
released mass on risk (reflected by the trends in V Ft F , Figure 159). Splitting the charge in half
through the use of two independent refrigerant circuits is also beneficial. In this case, the risk is
calculated as usual but with half the refrigerant charge, and the final result is doubled to account
for the two systems (circuits) existing concurrently. In this case, the risk associated with a smaller
charge in two separate circuits is less than a single system with twice the charge, indicating that
the contribution of charge size to the risk is greater than that of the increased f leak associated
with having two systems.
398
The effects of both hr and hd on risk also follow the behaviour of V Ft F , shown earlier in
Section 3.4.8. A lower hr increases risk but simultaneously lowering hd does not reverse this, as
implied by Figure 163. This is due to the influence o f large V Ft F occurring during off-cycle; if
on-cvcle were longer the benefit of lowering hd would become more apparent since a
correspondingly larger proportion of on-cycle V Ft F would be accounted for. This dependency on

cycling modes is equally relevant when unit airflow is doubled. Assuming a system operating
constantly, a doubling of the associated airflow would lead to a much lower V Ft F (as Figure
161), which would translate to a significant reduction in risk. However, with the system in off-
cycle, an absence of forced airflow can produce V Ft F that is typically several times larger than
the difference between V Ft F for low and high airflow rates. Thus, unless the contribution from
off-cycle V Ft F is reduced by a small time fraction (minimal off period), it would dwarf the
benefit offered by a higher airflow during on-cycle.
Similarly, it is seen in Table 52 that introducing a post-ventilation time ( tdmm, as calculated in
Section 3.3) had negligible effect on reducing risk. This is due to the fact that the risk during off-
cycle is unaffected by post-ventilation, and during on-cycle the post-ventilation only provides
additional airflow for (in this example) 300 s following the start of the release. Without post
ventilation, it was stated previously that the fans continue to operate until 50% o f the charge is
lost. Reference to Section 3.3.2 and 3.3.3 shows that the rate of floor concentration ( Cf )
development for large leaks is highest during the start of the release, whilst for longer leaks
Cf (t) —►CRm(t) . As such for catastrophic leaks, the effect of continued airflow for the first
50% of tIeak mitigates against the development of much high Cf anyw ay, and so the additional
post-ventilation will provide little extra benefit. For longer leaks, Cf is low anyway. Assuming
the risk is proportional to V Ft F , the insignificant benefit associated with post-ventilation is

reflected in Figure 162. Alternatively, where the unit airflow is initiated (or allowed to continue)
by the pressure switch or (more effectively) the refrigerant detector, a notable reduction in risk is
evident, and this is particularly evident when recalling that these control schemes introduce forced
airflow in both on- and off-cycle mode, the latter receiving most benefit in terms o f a greater
reduction in V F(F and therefore risk.
399
Concerning the integrity of the system, a reduction in the number of connections leads to a lower
risk and this is purely due to lower leak frequency. In terms of the integrity of equipment
electrical components (SOIs), two examples were given where the electrics are completely
unprotected ( Pamil = 1) - the first is with the SOI at unit height (2 m), and the second where the
SOI are located lower down (1 m). Whilst the presence of additional SOI would imply an
increase in risk, the results from the first case (unit SOI at 2 m) show negligible difference
compared to the reference example. This is on account of the positioning of the SOI being above
the region occupied by the flammable mixture. Conversely, with the second case where SOI are
closer to the floor, the effect was to significantly increase risk.
Factors other than those related to the equipment design also have significant effects on risk. It is
seen that room size has major influence, which again is implied by previous results for V Ft F
(Figure 160). The use of mechanical ventilation also reduces risk, but positioning of the inlet duct
only has minor influence. The amount of room fabric leakage (or A 50), also dictates the risk
because of its effect on dispersion in terms of airspeed, regardless of the operating mode.
Existence of thermal sources also has a similar benefit. Increasing the active time of room SOI
produced a slight increase in risk, and more so when the number of room SOI events doubled.
Other parameters not included in the evaluation can be commented on according to the
formulation of the model. In particular, if all components were within the room ( Pleak = 1), then
the risk would increase proportionally from the risk for Pleak =0.21 (i.e., 1/0.21). The same logic
applies to a change of system leak probability ( Psys), and the number of SOI (not on the
equipment). A lower fault probability or repair time for either unit airflow for mechanical
ventilation will obviously lead to a lower risk. Similarly, initiating airflow from system pressures
that correspond to a smaller leaked mass, or the detector set to a lower concentration would also
benefit risk reduction.
Based on this exercise, general guidance for risk reduction was developed. On the assumption
that a manufacturer cannot dictate the environment within which the equipment is installed, the
following are recommendations for design:
- Refrigerant selection for density close to air, and a high LFL (e.g., R1270)
- Small refrigerant charge
- Multiple independent circuits
- Piping and components located at highest position, and close (or above) air discharge
400
- High airflow rate and small discharge duct area
- Air discharge from floor level, or if at high level, discharge directed downwards
- Minimum number of connections, and the majority to be positioned w ithin outdoor unit
(if applicable)
- Unit SOI with good protection against ingress of gas, and positioned as high as possible
- Airflow1initiated with pressure transducer but preferably with a refrigerant detector
5.4 Example calculations for selected RAC equipment

To demonstrate the intended use of the QRA, four "‘case'5studies were prepared. Initial input
conditions were based on the minimum required by EN 378: 2000, and then re-evaluated with
certain (practicable) changes to the equipment design intended to improve safety. Since equipment
design does not affect the room characteristics and environmental conditions, these w’ere not
changed. In addition, certain equipment characteristics were also fixed; specifically, it was
assumed that the refrigerant charge was already minimised, the selected refrigerant was used on
the basis that it was thermodynamically preferable, and using multiple circuits would be too
costly. For all applications, room temperature were 21°C, average output of thermal sources was
150 W and their height was 1.5 m. Retention of refrigerant was 0.18 for on-cycle and 0.3 for off-
cycle, and during on-cycle, airflow' ceased once 50% o f the charge had been lost. Unit airflow
fault probability was 0.10, the repair time was 7 days, and no mechanical ventilation was used.
All SOI, both on the unit and within the room were 1 s, and Pm,ail was 0.14, and 1, respectively.
Window linear density, Wvp = 10 kg m"2 and the fixing force was equivalent to p vpfix = 1 kPa.
Values for other input parameters for the four applications are provided in Table 53, along with
the results from the QRA model. Variables adjusted to improve safety' are identified in square
parentheses, and similarly for the corresponding results.
Table 53: Reference values and variations
Integral chiller Split freezer Portable A/C in Split A/C in

Parameter
cabinet in shop cabinet in shop office office
Refrigerant R600a R290 R290 R1270
mr 0.5 kg 0.8 kg 0.4 kg 1.5 kg
L o tio n , Pleak 1 0.281 1 0.281
Leakage, P ^ 0.05 0.20 0.05 0.15
K 0.5 m [1 m] 0.5 m 0.5 m 2.5 m
Unit hd - [ l mj 2.0 m [0.5 ml 0.5 m 2.5 m
401
Integral chiller Split freezer Portable A/C in Split A/C in
Parameter
cabinet in shop cabinet in shop office office
Unit airflow 6 horiz. horiz. upward horiz. [down]
Unit Vd * - [0.08 m3 s'1] 0.12 m3 s'1 0.06 m3 s'1 0.23 m3 s'1
0.1 m2 [0.06
Unit A d - [0.2 m2] 0.6 m2 [0.3 in2] 0.4 m2 [0.3 m2]
m2]
N conn 20 [15] 50 [40] 20 [15] 50 [40]
Unit N soi 10 20 [15] 10 15 [5]
Unit N e 4 per 24 h 4 per 24 h 4 per 24 h 4 per 24 h
K m * A Rm 3 m x 30 m2 3 m x 30 m2 3 m x 20 m2 3 m x 40 m2
N soi (low. mid. high) 20, 5, 5 20. 5, 5 10. 2, 2 20, 2, 2
N g (low. mid. high) 8, 4, 4 per 24 h 8, 4, 4 per 24 h 4, 2, 2 per 24 h 4, 2. 2 per 24 h
Window A v 3-> in2 3 m2 2 m2 4 m2
On-cycle top 18 h per 24 h 18 h per 24 h 8 h per 24 h 8 h per 24 h
Loss of m r to stop Vd 50% 50% 50% 50%
Hocc 8 8 3 6
Occupancy tocc 12 h per 24 h 12 h per 24 h 8 h per 24 h 8 h per 24 h
Post-ventilation td - [240 s] - - [240 s] -
C detect t0 start Vd - - [20% LFL] - -
Loss of mr to start Vd - - - - [25%]
A 50 20 h*1 20 h'1 10 h 1 10 h'1
Thermal source ttherm 12 h per 24 h 12 h per 24 h 8 h per 24 h 8 h per 24 h
N therm 10 10 9 18
2.0* 10'9 3.6xl0'8 2.3 x 10'10 2.4xl0'9

f (y 1)
[2.6xlO'10] [7.6x1 O'9] [1.4xlO'10] [1.6x10'*°]
Max I occ (s (kW m'

2)4 ,3 ) 0.9 [0.8] 2.2 [2.2] 1.6 [1.6] 1.4 [0.1]
MaX ty°Rm 0 ^ ) 8.6 [5.9] 26.6 [26.6] 4.9 [4.9] 8.5 [0.4]
2.1x 10'9 8.9xl0'8 2.3 xlO'10 3.6X10'9

X (-)
[2.0xlO'10] [1.6xl0'8] [1.6xlO-10] [1.6x10-**]
t Given in volume flow rate instead of air changes because they are equipment specific.
402
The integral chiller cabinet used natural convection for the condenser, so forced airflow was
introduced by employing a condenser fan. For a small unit, using a pressure sensor or refrigerant
detector is too costly, so post-ventilation was used to continue airflow for an additional 240 s.
Piping connections w ere reduced and the height of the low est refrigerant containing part was
raised to 1 m by rearranging pipework. The results indicate that these measures could lessen the
risk (in terms of ignition frequency and risk factor) by a factor of 10, and the maximum severity
of the consequences was also reduced slightly.
For the split freezer cabinet, the condenser air discharge was positioned much closer to the floor,
and the discharge velocity7was increased by halving the outlet duct area. A refrigerant sensor was
positioned at floor level, which initiated the condenser air (if not already operating) if the
concentration exceeded 20% of the LFL. Additionally, the number of connections and SOI were
reduced by locating those that could be, to the outside unit. The effectiveness of these measures
was slightly less than with the chiller cabinet, where both ignition frequency and risk factor were
about five times lower. The maximum consequence was unchanged.
In the case of the portable air conditioner, the amount of possible alterations is limited because of
compromising its ease of use, for example, a floor based appliance is easier to manoeuvre than a
ceiling mounted unit. Other than modifying the piping, only the air discharge duct was made
smaller and post-ventilation was used. Consequently, only a minor reduction in ignition frequency
and risk factor occurred, where the altered equipment lias about half that of the original design,
and similarly, there was virtually no change in the maximum overpressure and thermal intensity'.
A similar approach to the split freezer cabinet was used for the split air conditioner, where
additional piping connections and SOI were located within the outdoor unit: Also, the air
discharge duct was given a smaller area whilst being directed downward, and in the case of a leak
during on- or off-cycle, a pressure sensor would initiate the airflow. With these additional
features, a reduction in ignition frequency by a factor of 15 was achieved whist the risk factor
was lowered by some 200 times. This is a result of the major reduction in severity of maximum
overpressure and thermal intensity too.
In addition to the exercise above, calculations were also made for servicing activities. For each
application, the values in Table 53 were used, except where interference from a service technician
would normally cause them to change. These included the variations listed below7.
- Cycling time is assumed to be split equally between on and off mode, i.e., <f>op = 0.5.
403
- Release height is at 0.5 m to account for leaks originating from service manifolds, hoses
and cylinders, in addition to the actual refrigerant circuit.
- Additional SOI for sendee tools, specifically a brazing torch (used twice for ten minutes),
sparker, and vacuum and recovery pumps.
- Some equipment electrical devices become unprotected thereby increasing availability of
SOI, so P ^ j = 0.43.
- Leak frequencies for both on- and off-cycle increase (as discussed in Section 2.3), for
which a factor of ten w as used, and off-cycle f leak was set as the value for on-cycle.
The risk calculations w ere based on a single two-hour service visit in one year, the only occupant
is the sendee person and SOIs elsewhere in the room are ignored. The results of the calculations
in terms of / * and X are provided in Table 54, for the four different (unadjusted) applications
detailed in Table 53. Also given are / * and X , relative to the respective values under normal
operation.
Table 54: ORA model results fo r servicing on fo u r applications
Ignition frequency Risk factor

Application / ‘ (sen1) / X (serv)/
f * (y1) / / ‘ (norm)
X (-) /X (norm )
Integral chiller
1.2*10'7 60 9.5*10'8 45
shop
Split freezer cabinet l.oxio-6 29 1.9X1 0 6 21
Portable A/C 7.0* 10'8 308 6 .6 X10"8 284
Split A/C 8.6* 10’7 352 1.4X10'6 378
In all of the examples, the risk is seen to be higher for servicing activities compared to normal
operation. This result is expected due to higher leak frequencies, more vulnerable SOI and larger
flammable volumes caused by releases (in some cases) occurring from positions closer to the
floor. For the chiller and freezer, the increased risk is relatively small compared to the air
conditioners, and this is explained as follow s. When the a/c is in on-cycle, there is usually a high
room air speed, which offsets the corresponding high on-cycle f leak by producing much smaller
flammable volumes, whilst both chiller and freezer produce no, or low very low air speeds in on-
cycle. As such, the result of increasing the chiller or freezer off-cycle f laok to equal that at on-
cycle is comparable to simply increasing the on-cycle time to (f>op ~ 1, since the resulting V Ft F
404
is about the same in either mode. However, the air conditioners produce a relatively large V FtF
during off-cycle which is normally offset with the low corresponding f kak, so increasing off-
cycle f leak greatly overwhelms the contribution of the on-cycle V Ft F . This effect is amplified in
the case of the split unit because of the lower release height.
In general, the differences between operating and servicing / * and X are consistent with the
review of incidents (Table 49) where very few apply to the operating mode. Nonetheless, by
examining the contribution of each different variable that contributes to a higher risk during
servicing can help identify preferred behaviour and use o f tools that would minimise the hazard.
1E+06 T 100%
80%
1E+04 dP
60% J 8
E
40%
1E+02 -
20%
0% n........................... . —ri_
1E+00
0 0.02 0.04 0.06 0.08 0.1
0.001 0.01 0.1 1
C/m a r^ n>3) Cfma* ^ n>3)
Figure 180: Relationship between Ckmax and Figure 181: Relationship between Cfmax and
total ( r = 0.82) the proportion o f V ^ f above 1 m
5.5 General risk dependency on maximum concentration

Intuitively, it is expected that higher Cf max leads to a greater risk. In order to address this, results
for C f max and subsequently / * were generated over a range o f input parameters so they could
be used to identify any general tendencies. First, the entire set o f V Ft F results (Figure 159 to
Figure 163) from the dispersion model was plotted against C f max to identify any general
relationship. The data is shown in Figure 180 where a scattered s-curve is observed that is
bisected by the LFL of R290. When Cfmax is above LFL (> 0.041 kg m'3) a high V Ft F always
occur (i.e., > 5000 m3 s), and below7the LFL (< 0.034 kg m'3) only small V Ft F occur (i.e., <
500 m3 s). This observation provides a useful indication that a small V Ft F is almost guaranteed
405
if the conditions lead to Cf max < LFL. This was consistent with the earlier recommendation that
equipment should be designed to maintain Cf max to below about LFL ^ 1.28 to avoid local
concentrations reaching LFL (Section 3.3).
1E-K)1
eo
corresponds to
LFL
IE-05 -
1&01 1E+01 1E+03 1E+05 1E+07 0 0.05 0.1 0.15
V Ft F (m3 s) C fm a x (kgm-3)
Figure 182: Correlation between P^* and Figure 183: Relationship between and
(r=l) Cfmax( r =0.82)
The influence of SOI positioning was considered since the probability calculations in Section 2.1
require the room be divided into three levels of equal volume to account for the locality' of SOI. In
Section 5.2 (Figure 174) it was seen that the greater proportion of / * shifted from the lower
level to the middle level as leak times got longer, which was commensurate with an overall
reduction of / * . Plotting the ratio of V Ft F existing within the middle level (above 1 m and
below 2 m) to total V Ft F indicated when it is important to consider SOI positioned above floor
level (Figure 181). For many situations a lower C fmax is consistent with a greater proportion of
V Ft F occurs at the higher level. Although it was found that the greater part of V Ft F and / *
did occur within the middle level under these conditions, these corresponded to very low Cfmax (<
LFL -s- 2). Such low Cf max were due to long leak times and/or high uRm , consequently resulting
in minute / * relative to / * from catastrophic leaks. However, under certain conditions there is
a contrary' tendency for an increasing proportion of V Ft F to occur in the middle level as Cftnax
rises. This is generally when Cf max < LFL and very' large flammable clouds develop that already
occupy the entire lower level.
406
P F* was determined from the same V Ft F and Cfmax data using other simple values: = 1,
P^s = 1, Pperc = °-73- fso, = 1 s, N sot = 1, and N e = 1. Figure 182 shows the linear
relationship between P F* and V Ft F , which was found to apply generally provided tsoi < 100 s,
N soi <30. Given the linearity7with V Ft F , P F* was justifiably plotted against Cf max (Figure
183). As with the observations of C fmax and V Ft F , similar generalisations are made: where
C fmax > LFL always P F* > 2 x 1C4, below LFL, P F* < 1* 1C4, and the range 1*10^ < P F* <
2* 10^ generally sits on the LFL. Assuming a constant f leak the trends in P F*are considered to
apply equally to / * .
r 1E+04 ----- —
* 1E+02 - ^ 1E+02 -
o Overpressure n Overpressure
o Thermal intensity <>Thermal intensity
1E+01 0 0.05 0.1 0.15
C g) C f
fm a x
(kgm-3)'
Figure 184: Relationship between mean Figure 185: Correlation betM'een Cfmax and
flammable mass and consequence (r2 = 0.99) consequence ( r = 0.83)
Trends in 1^ and ApRm with m F as the selected flammable quantity were plotted in Figure
184, and these also produced a linear relationship; this is expected since both 1occ and t±p°Rm are
directly dependent upon the available energy within a given mass of mixture. As with P F*, the
relationship between I ^ and ApRm and Cf max w?as plotted (Figure 185). It is seen that below
the LFL, the consequence is often insignificant (< 1 kPa and < 1 s (kW m'2)j/4), yet beginning to
exceed 50 kPa and 100 s (kW rrf2)3'4 at C f max slightly above the LFL.
407
SECTION 6: CONCLUSIONS
6.1 Summary of work

Observations from a literature review were used to identify the tasks necessary to enable a more
comprehensive QRA method of RAC equipment; more precise leak characterisation, probability
calculations for localised conditions, behaviour of the dispersion process of refrigerants leaks
(focussing on the influence of airflows), consequence analysis, and criteria forjudging the
acceptability of risk. Initially, a hazard analysis was used to derive a methodology for calculating
ignition and consequence frequencies, and an acceptability7criteria based on the chosen risk
measures. Methods were derived for characterising the likelihood of leakage, active sources of
ignition and flammable concentrations. The latter were based on characterisation of room
concentrations from a refrigerant leak, for which empirical correlations were developed using
measurements from a series o f experiments. In the event o f ignition, consequences were quantified
by modelling thermal intensity and room overpressures.
Overall risk of RAC equipment was found to be influenced by a large number of variables and
the QRA accounts for around 50 of these, which can be adjusted according to the design,
construction and behaviour of the equipment and the local environment. The QRA model is
broadly generic within the range of characteristics associated with RAC equipment normally used
in domestic and commercial contexts provided the conditions are within the applicability7limits of
the sub-models. Examining QRA model outputs led to identification of design features that can
minimise the risk associated with RAC equipment.
6.2 Key findings and originality

Several important findings were made, particularly with respect to the dispersion of a leak of
flammable refrigerant. For any specific application, no single design feature can be practically
used to guarantee low risk because other combinations of factors may result in the opposite.
How ever, the QRA model did indicate a number o f techniques - preferably used in combination -
that can contribute to improved safety7. The risk associated with equipment already conforming to
safety standards can vary widely, and checking various design features can lead to a marked
reduction in risk, by a factor o f up to 102 or more.
Although not exactly comparable, the QRA model was used to “re-calculate” published QRAs
which use the conventional approach; under most situations the re-calculated risk w7as
significantly lower by up to a factor of 103. This w7as due to the current method having greater
408
precision on account of the wide use of empirical data and discretisation of conditions, thereby
avoiding the compounding effect of numerous worse case assumptions. Also, the published QRAs
presented results typically for one or two fixed set of environmental conditions thereby negating
any anticipated changes over time. Moreover, certain combinations o f equipment and
environmental characteristics could potentially create an unexpectedly high risk situation, and by
adopting very general assumptions, and neglecting particular influences, a high risk situation
could be overlooked.
Throughout the preparation of this study, there have been several important developments.
- Evaluation o f risk. Several important measures of the “risk” were identified: frequency of
ignition, frequency of fatality, maximum thermal intensity and overpressure, and additionally,
the newly developed measures o f integrated thermal intensity-frequency, integrated
overpressure-ffequency, and in combination, a generalised “risk factor”. The calculation of
these risk measures was based directly on leakage characteristics, flammable concentration
and SOI probabilities, and time fractions of transient conditions. The criteria for establishing
whether results constituted a “safe” or “unsafe” risk were determined from knowledge of the
risk associated with equipment having socially “acceptable” safety' record, and tolerances
specified by the national safety authorities, for example an ignition frequency lower than 1(T6
y'1 are considered to pose a “negligible” risk .
- Leak frequency. Existing QRAs have employed anecdotal data on leak frequency and
normally categorised in terms of “catastrophic” leak and a “small” leak. As part of this study,
data for released quantities and leaks duration was made available. This data was employed
to develop a method for calculating leak frequencies and durations over a wide range of leak
sizes as a function of the number of connections or joints associated with a system.
- Probability models. Existing QRA calculate ignition probabilities based on large-scale
outdoor releases. Instead, an approach was developed to account precisely for all known SOI
associated with the RAC equipment and others within a given room.
- Concentrations from leakage. A number of full-scale dispersion experiments were
conducted, allowing the concentration development to be studied, whilst accounting for the
influence of forced airflow and other parameters normally associated with RAC equipment.
For example, by increasing mean air speed from 0.05 m s'1to 0.2 m s'1 can reduce maximum
floor concentrations from twice the LFL to 0.25 * LFL. Empirical correlations were
developed to estimate concentrations arising from catastrophic leaks (where the entire charge
is released in 3 minutes) and releases over longer durations. Complimentary to their
application in the dispersion model, these formulae can be employed for evaluating the design
409
of RAC equipment in terms of maximum concentrations resulting from leaks? used
iteratively to establish equipment design that avoids flammable floor concentrations.
- Modelling dispersion o f a leak. Quantification of flammable volume, flanunable mass and
flammable time was achieved through a numerical model that addressed the descending
plume stage, spreading plume and decaying cloud parts.
- Calculation o f severity o f consequences. Consequences of ignition were identified as thermal
intensity and room overpressure. In particular, the explosion process (leading to
overpressure) was evaluated with a newly developed model of burning o f a partially
flammable cloud in a confined and/or vented space. Comparison with eSper^menta^
showed good agreement, typically within ±100% o f the maximum overpr6Ssure-
- Determination o f risk. The influence of variables associated with equip*11^ design, external
conditions and operation, on the risk measures was assessed with the m°del. The
observations w ere used to establish general guidance for the design of equipment to nunimise
the risk, providing additional requirements to those specified in the safety standards.
Whilst the significance of the contribution offered by this study is largely siibjecdve? considering
the scale of the related applications implies a potentially far-reaching influef,ce- International
agreements on ozone depleting and global warming gases (typically fiuoroc#rh°ns) mean that
many refrigerants used historically or indeed recently introduced, are b e c o m e s increasingly
restricted. One option for RAC equipment manufacturers is to build system5 that use
environmentally benign gases, which include HCs (UNEP, 2002). Currently ? ah°ut one-quarter of
new domestic refrigerators produced globally use R600a. Within Europe an increasing number o f
producers of a variety of small and medium-sized commercial cabinets have adopted R290, and
manufacturers of split and integral air conditioners offer systems with R290 ^ d ^1270. As
regional legislation further restricts fluorinated refrigerants, it is likely that tf ie uptake of HCs will
accelerate. Furthermore, developing countries and countries in economic t r a ^ ^ io n are adopting
HCs (in addition to other alternative refrigerants). For domestic and commefc ‘a^ sectors,
estimates based on UNEP (2002) suggest around 150 million appliances a r^ produced globally
each year, and some 20% of these may soon be using flammable refrigerants - ^ *s ev^ ent horn
these numbers that a significant reduction in potential incidents could be ach»*eved by using a
more comprehensive approach to safe design.

Use of the QRA model is limited to the range of conditions that can be reliab ly accounted for and
the accuracy that the variables can be represented, as well as correct in te r p r ^ ^ o n of input
410
of RAC equipment in terms of maximum concentrations resulting from leaks, and used
iteratively to establish equipment design that avoids flammable floor concentrations.
- Modelling dispersion o f a leak. Quantification of flammable volume, flammable mass and
flammable time was achieved through a numerical model that addressed the descending
plume stage, spreading plume and decaying cloud parts.
- Calculation o f severity o f consequences. Consequences of ignition were identified as thermal
intensity and room overpressure. In particular, the explosion process (leading to
overpressure) was evaluated with a newly developed model of burning of a partially
flammable cloud in a confined and/or vented space. Comparison with experimental data
showed good agreement, typically within ±100% o f the maximum overpressure.
- Determination o f risk. The influence o f variables associated with equipment design, external
conditions and operation, on the risk measures was assessed with the QRA model. The
observ ations were used to establish general guidance for the design o f equipment to minimise
the risk, providing additional requirements to those specified in the safety' standards.
Whilst the significance of the contribution offered by this study is largely subjective, considering
the scale of the related applications implies a potentially far-reaching influence. International
agreements on ozone depleting and global warming gases (typically fluorocarbons) mean that
many refrigerants used historically or indeed recently introduced, are becoming increasingly
restricted. One option for RAC equipment manufacturers is to build systems that use
environmentally benign gases, which include HCs (UNEP, 2002). Currently, about one-quarter of
new domestic refrigerators produced globally use R600a. Within Europe an increasing number of
producers of a variety of small and medium-sized commercial cabinets have adopted R290, and
manufacturers of split and integral air conditioners offer systems with R290 and R1270. As
regional legislation further restricts fluorinated refrigerants, it is likely that the uptake of HCs will
accelerate. Furthermore, developing countries and countries in economic transition are adopting
HCs (in addition to other alternative refrigerants). For domestic and commercial RAC sectors,
estimates based on UNEP (2002) suggest around 150 million appliances are produced globally
each year, and some 20% o f these may soon be using flammable refrigerants. It is evident from
these numbers that a significant reduction in potential incidents could be achieved by using a
more comprehensive approach to safe design.
6.3 Limitations and further w ork

Use of the QRA model is limited to the range of conditions that can be reliably accounted for and
the accuracy that the variables can be represented, as well as correct interpretation of input
410
values. As with any numerical model, extending the range of applicability and raising “apparent”
accuracy can be achieved through improving sub-models, and their integration into the overall
model. Similarly, wherever empirical correlations or data have been employed, there is always
scope for improvement with additional data, and for a wider range of variables.
The following list suggests options for improvements.

- RAC equipment construction. Currently, the QRA can handle a limited number of design and
construction features associated with equipment. Some equipment differs from the generic
representation established earlier, such as when the entire system is located within housing,
where piping and components are routed around a room, and where there are multiple or
shifting air discharge ducts. Therefore, a broader range of equipment characteristics could be
accounted for if more comprehensive correlations were developed from new measurements,
and the appropriate input options provided.
- Probability calculations. It is possible to develop more sophisticated probability models,
which could be employed where variables that impact on the behaviour of leakage,
dispersion, and activation of SOI simultaneously are linked (such as occupancy, on-cycle
times and presence of SOIs). Further, ageing of equipment will increase the frequency of
leakage, mechanical faults and integrity of SOI protection. This aspect has not been
addressed, although given that such failures are known to change with time, risk is likely to
increase over equipment lifetime. A proper assessment of ageing would again be drawn from
empirical data. Also, a better understanding of the likelihood of flammable concentrations
reaching SOI within their casing or housing w ould also be valuable, rather than assuming
flammable concentrations are always immediately available.
- Release characteristics. Currently, leaks are assumed as constant mass flow rate, although it
is known that leaks often occur intermittently and variation of conditions can lead to changes
in the leak aperture and subsequently mass flow. Better characterisation of leak patterns is
needed, and how equipment operation behaves in response to leakage (e.g., how release rates
vary and the relationship between charge deficit, pressures and system controls). Also
leakage is known to be influenced by quality o f materials and workmanship, which could also
be characterised.
- Dispersion modelling. Additional experimental data is required for effects of leaks from
various equipment components (rather than a diffuser for w'orse-case). More comprehensive
data and modelling are required on dispersion behaviour due to the effect of multiple sources
of airflow occurring simultaneously within the room. Airflows from RAC equipment are
normally at a higher or lower temperature than the ambient, and it is likely that the resultant
411
difference in air buoyancy affects dispersion characteristics, also demanding more
comprehensive evaluation of infiltration for rooms within buildings.
- Uncertainty analysis. It would be beneficial in terms of interpretation of results to include
some indication of the uncertainty associated with each part of the QRA, and accordingly, the
final risk measures.
- Consequence. Ignition within either RAC equipment housing or other SOI-containing
enclosures can produce high overpressures, shock waves and flying fragments; including
these elements would improve consequence analysis. Depending upon its geometry, ignition
may be extinguished if the flame front is prevented from reaching a larger flammable mixture
by small apertures. Physical examination of SOI-containing equipment and associated
measurements would be needed for this.
- General validation. It would be useful to conduct a series of experiments where, under
different equipment/room configurations, a “realistic” release is made and potential SOI are
activated. The measurements for all intermediate parameters (such as release rates, air
speeds, concentrations, overpressures, heat flux, etc) can be compared against the output of
the various sub-models, ultimately to help validate the overall QRA model. Additionally, the
model could be adapted to permit inputting of measurement data from leak simulation tests
(as obliged by certain safety standards) so that dispersion modelling could be calibrated for
the equipment under consideration.
Lastly, of interest to equipment producers, is the capital and operational costs of including certain
design features that could reduce risk. Therefore, addressing this cost effectiveness would provide
useful and practical information.
412
REFERENCES
ACRIB. Guidelines for the safe use of hydrocarbon refrigerants. Air Conditioning and
Refrigeration Industry Board. London UK, 2000.
ADL. Risk assessments of flammable refrigerants. Report for Calor Gas Ltd. Arthur D. Little,
Cambridge, UK. 1998.
AQei, T. Swiss proposal. P028, Proc. IEC/ISO JWG, Paris, February 1999.
Aloisi, A. Personal communication. Experimental tests on propane leaks. De’Longhi SpA,

Treviso, Italy, 17th May 1999.
Aloisi, A. Personal communication. Experimental tests on propane leaks effects of appliance

ventilation. De'Longhi SpA, Treviso, Italy, 20th September 1999.
Aloisi, A. Personal communication. Experimental tests on propane leaks effects of appliance

ventilation and leak time. De'Longhi SpA, Treviso, Italy, 20th February 2000.
Anthony, E. J. The use of venting formulae in the design and protection of building and industrial
plant from damage by gas or vapour explosions. J. Hazardous Materials, Vol. 2, pp. 23-49.
1997.
Ayers, J. Parasense study on refrigerant leaks. Parasense Environmental Products Ltd,

Gloucester, UK. November 2000.
Babrauskas, V., Peacock, R. D., Reneke, P. A. Defining flashover for fire hazard calculations:
Part II. Fire Safety Journal, Vol. 38, pp. 613 - 622. 2003.
Baines, W.D., Turner, J. S. Turbulent buoyant convection from a source in a confined region. J.
Fluid Mech., vol. 37, part 1, pp. 51-80, 1969.
Baines, W.D., Turner, J. S., Campbell, I. H. Turbulent fountains in an open chamber. J. Fluid
Mech., vol. 212, pp. 557-592, 1990.
Barnett, S. On buoyant gas mixing in confined regions. PhD Thesis, University of Cambridge,
1991.
Barry, T. F. Application of fire and explosions risk assessment to an LPG bulk storage facility.
Fire Hazard and risk Assessment, ASTM STP 1150, M. M. Hieschler (Ed), pp. 183 - 206. 1992.
Beyler, C. L. Fire hazard calculations for large, open hydrocarbon fires. In SFPE Handbook of
Fire Protection Engineering, NFPA, USA. 2002.
413
Blackwell, N., Bendixen, L., Birgfeld, E. Risk assessment og HFC 152a and carbon dioxide
mobile air conditioning systems. Proc. Earth technologies forum, Washington, USA. April 2004.
Blades, A., Colboume, D. Report on the investigation of an incident involving a domestic

refrigerator using R600a. Association of Manufacturers of Domestic Electrical Appliances
(AMDEA), London, UK. 2004.
Blanken, A. W. T. van, Verwoerd, M. Quantitative risk analysis of an ice cream cabinet using
hydrocarbon refrigerant. R2001/476. Report of TNO, Apeldoom, The Netherlands. December
2001 .
Blom-Bruggeman, J. M, van Gerwen, R. J. M., Verwoerd, M. Risk Assessment of typical cooling

and heating systems using natural working fluids. Part: Risk assessment methodology. R96-348.
Report of TNO, Apeldoom. Netherlands. 1996a.
Blom-Bruggeman, J. M, van Gerwen, R. J. M., Verwoerd, M. Risk Assessment of a Bulk Milk

Tank using Hydrocarbon Refrigerant. Final Report o f TNO (for Novem B.V.), Apeldoom.
Netherlands. 1996b.
Britter, R. E. A review of some mixing experiments relevant to dense gas dispersion. From
‘‘Stably Stratified Flow and Dense Gas Dispersion” by J. S. Puttock (Ed.), Oxford University
Press. UK, 1988.
Brown, M. J., Palarya, S., Snyder, W. H. Plume descriptors derived from non-gaussian
concentration model. Atmospheric Environment, Vol. 31, No. 2, pp. 183-189. 1997.
BS 5925: 1991 Code of Practice for Ventilation Principles and designing for natural ventilation.
British Standards Institution, London, UK. 1991.
Centre for Chemical Process Safety (CCPS). Use of vapour cloud dispersion models. American
Institute of Chemical Engineers, New York, USA, 1996.
Cleaver, P., Personal communication, BG Technology7Pic., Loughborough, UK, 1999.
Cleaver, R. P., Marshall, M. R., Linden, P. F. The build-up of concentration within a single
enclosed volume following a release of natural gas. J. Hazardous Materials, no. 36, pp. 209-226,
1994.
Clodic, D., Cai, W. Study of diffusion o f propane and iso-butane in rooms respecting the
standardised practical limit. Proc. ASHRAE Annual Meeting, Boston, USA, 1997.
Clodic, D., Cai, W. Tests and simulations of diffusion of various hydrocarbons in rooms from air
conditioners and refrigerators. Proc. IIR Conf. Natural Working Fluids, Aarhus, Denmark, 1996.
414
Clodic, D. Leak flow rates and measurements of concentration gradients of flammable
refrigerants. A1CARR Conf. “FREE '9 T \ Verona, Italy. April 1997.
Colboume, D., Suen, K. O. Appraising the Flammability Hazards of Hydrocarbon Refrigerants

using Quantitative Risk Assessment Model. Part I: Modelling Approach. Int. J. Refrig., pp. 774 -
783, Vol. 27. 2004a.
Colboume, D., Suen, K. 0 . Appraising the Flammability Hazards of Hydrocarbon Refrigerants

using Quantitative Risk Assessment Model. Part II: Model evaluation and analysis. Int. J.
Refrig., pp. 784 - 793, Vol. 27. 2004b.
Colboume, D., Suen, K. O. Equipment design and installation features to disperse refrigerant
releases in rooms. Part I: Experiments and analysis. Int. J. Refrig., No. 6, Vol. 26. 2003a.
Colboume, D., Suen, K. O. Equipment Design and Installation Features to Disperse Refrigerant
Releases in Rooms. Part II: Determination of procedures. Int. J. Refrig., No. 6, Vol. 26. 2003b.
Colboume, D., Suen, K. O., Ritter, T. J. A Risk Assessment Model for Hydrocarbon
Refrigerants. Proc. Inst. Refrig., London, 2003.
Collier, P., Watson, L. Fire Statistics United Kingdom 1997. Home Office Statistical Bulletin,
Research Development and Statistics Directorate, London, UK. November 1998.
Committee for the Prevention of Disasters (CPD). Guidelines for quantitative risk assessment of
hazardous materials (liquids and gases). CPR-18E (Purple Book), 1st edition. The Hague, The
Netherlands. 1999.
Committee for the Prevention of Disasters (CPD). Methods for the calculation of the physical
effects due to releases of hazardous materials (liquids and gases). CPR-14E (Yellow Book), 3rd
edition. The Hague, The Netherlands. 1997.
Committee for the Prevention of Disasters (CPD). Methods for the calculation of consequences of
the releases of hazardous materials (liquids and gases). CPR-16E (Green Book), 1st edition. The
Hague, The Netherlands. 1992.
Committee for the Prevention of Disasters (CPD). Methods for determining and processing
probabilities. CPR-12E (Red Book), 1st edition. The Hague, The Netherlands. 1988.
Crowl, D. A. Understanding explosions. Centre for Chemical Process Safety (CCPS). American
Inst. Chem. Eng., New York. 2003.
Davies, P., Pearson, J. Gas In relations to other domestic risks. Business Strategy' Group
Research project for the Health and Safety Executive, Final Report. UK, August 1999.
415
DeHaan, J. D., Crowhurst, D., Hoare, D., Bensilum, M., Shipp, M. P. Deflagrations involving
stratified heavier-than-air vapour/air mixtures. Fire Safety J., Vol. 36. 2001.
Ditali, S., Colombi, M., Moreschini, G., Senni, S. Consequence analysis in LPG installation
using an integrated computer package. J. Hazardous Materials, Vol. 71, pp. 159 - 177. 2000.
Douglas, J. F. Gasiorek, J. M., Swaffield, J. A. Fluid Mechanics. Longman Scientific and

Technical. 1986.
Duijm, N. J., Ott, S, Nielsen, M. An evaluation of validation procedures and test parameters for
dense gas dispersion models. J Loss Prevention in the Process Industries, Vol. 9, No. 5, pp. 323 -
338. 1996.
Elbers, S. J., Verwoerd, M. Quantitative Risk Assessment of a Heat Pump System with Propane
Refrigerant. Final Report of TNO, R97-238. Apeldoom, The Netherlands. 1997.
EN 378: Parts 1 - 4 . Refrigerating systems and heat pumps - Safety and environmental
requirements. BSI, London. 2000.
EN 60335-2-24: 2000. Household and similar electrical appliances - Safety - Part 2-24.
Particular requirements for electric refrigerators, freezers and ice-makers. BSI, London, 2000.
EN 60335-2-89: 2004. Household and similar electrical appliances - Safety - Part 2-89:
Particular requirements for commercial refrigerating appliances with an incorporated or remote
refrigerant condensing unit or compressor. BSI, London, 2004.
EN TR 14739: 2004. Scheme for carrying out a risk assessment for flammable refrigerants in
case of household refrigerators and freezers. BSI, London. 2004.
Etheridge, D., Sandberg, M. Building ventilation: theory and measurement. British Gas Research
and Technology, John Wiley & Sons, UK, 1996.
Fletcher, B., Johnson, A. E. The accumulation of gases in ventilated and unventilated enclosures.
Ventilation '85, H. D. Goodfellow (Ed.), Amsterdam, 1986.
Gerwen, R. J. M., van, Jansen , C. M. A. Risk assessment of flammable refrigerants. Proc. IIR
Conf. Natural Working Fluids, Hannover, Germany. 1994.
Gerwen, R. J. M., van, Verwoerd, M. Quantification and evaluation of safety risks related to the
use of ammonia and hydrocarbons as refrigerants. Proc. IEA Heat Pump Programme Annex 22
Workshop “Compression systems with natural working fluids”, HPP-AN22-3. October 1996.
Gerwen, R. J. M., van, Koffijberg, H. Natural refrigerants: applications and risks. Proc. IIR 19th
Int. Congr. Refrig. The Hague, The Netherlands. 1995.
Goetzler, W., Bendixen, L., Bartholomew, P. Risk assessment of HFC-32 and HFC-32/134a
(30/70 wt. %) in split system residential heat pumps. Final Report DOE/CE/23810-92 for ARTI,
Arthur D. Little, USA. April 1998.
Gopalakrishnan, S. G., Sharan, M. A lagrangian particle model for marginally heavy gas
dispersion. Atmospheric environment, Vol. 31, No. 20, pp. 3369 - 3382. 1997.
Guvoncourt D. M. M., Fennell, P. H. Potential catastrophic failures o f refrigerant systems. AEA

Technology Report, Didcot, UK, November 2000.
Harris, R. J. Gas explosions in buildings and heating plant. British Gas Corp., EF & N Spon
Ltd., UK, 1983.
Health and Safety Executive (HSE). Reducing risks, protecting people. HMSO, Norwich, 2001.
Heiselberg, P. Concentration distribution in a ventilated room under isothermal conditions. Proc.

12th AIVC Conf., Air movement and ventilation control within buildings, Ottowa, Canada, 1991.
Hjertager, B. H., Bjorkhaug, M., Fuhre, K. Explosion propagation on non-homogenous methane-

air clouds inside an obstructed 50 m3 vented vessel. J. Hazardous Materials, Vol. 19, pp. 139 -
153. 1988.
Hochst, S., Leuckel, W. On the effect of venting large vessels with mass inert panels. J. Loss
Prevention in the Process Industries, Vol. 11, pp. 89 - 97. 1998.
Hvmes, I., Boydell, W., Prescott, B. Thermal radiation: Physiological and pathological effects.
Institution of Chemical Engineers, Rugby. 1996.
Institute of Refrigeration (IOR). Code of Practice for the design on refrigeration systems using
A3 refrigerants. Institute of Refrigeration, UK, 2000.
Jabbour, T. Clodic, D. Ignition tests of flammable refrigerant leaks in ventilated and unventilated
rooms. Proc. 5th IIR Gustav-Lorentzen Conf., Guangzhou, China. 2002.
Jabbour, T., Loyer, J. C., Clodic, D. Analysis of the IEC requirements for the use of flammable
refrigerants. Report of Centre d'Energetique, Ecole des Mines, Paris. 2003.
Jansen, C. M. A., van Gerwen, R. J. M. Risk assessment of the use of flammable refrigerants. (In
transport refrigeration.) R95-189. Final Report of TNO, Apeldoom, Netherlands. 1996.
Kataoka, O. Personal communication. Interim dispersion analysis report. Daikin Industries,

Osaka, Japan. October 1999.
Kataoka, O. Background information for CDV formulas for flammable refrigerants. Proc. ISO
TC 86 SC 1/IEC SC 61 D Joint Working Group. 2001.
417
Kataoka, O., Yoshizawa, M., Hirakawa, T. Allowable charge calculation method for flammable
refrigerants. Proc. Int. Refrig. Conf. Purdue, USA. 2000a.
Kataoka, 0 ., Ishida, S., Hirakawa, T., Yoshizawa, M. Experimental and numerical analyses of
refrigerant leaks in a closed room. ASHRAE Transactions, pp. 1151 - 1158, ASHRAE, Atlanta,
USA. 2000b.
Khan, F. I., Abbasi, S. A. TORAP - a new tool for conducting rapid risk assessment in
petroleum refineries and petrochemical industries. J. Loss Prevention in the Process Industries,
Vol. 12, pp. 299 - 313. 1999. (Also in Applied Energy, Vol. 65, pp. 187 - 210. 2000.)
Kumar, A., Mahurkar, A., Joshi, A. Study o f the spread of a cold instantaneous heavy gas release
with surface heat transfer and variable entrainment. Journal of Hazardous Materials, Vol. 101,
pp. 1 5 7 - 177. 2003.
Leach, S. J., Bloomfield, D. P. Ventilation in relation to toxic and flammable gases in buildings.
Build. Sci. Vol. 8., pp. 289-310, 1973.
Lee, K-W. A methodology for assessing risk from released hydrocarbon in an enclosed area. J.
Loss Prevention in the Process Industries, Vol. 15, pp. 11 - 17. 2002.
Manager, H. Personal communication. Danfoss, Flensberg, Germany, 2001
Marshall, M. R. The effect of ventilation on the accumulation and dispersal of hazardous gases.
Proc. Inst. Chem Eng. 4th Int. Symp. On loss prevention and safety promotion in the process
industries, Harrogate, Series No. 82, 1984.
McLinden, M. O., Klein, S. A., Lemmon, E. W., Peskin, A. P. NIST Thermodynamic Properties
of Refrigerants and Refrigerant Mixtures Database (Refprop). Version 6.01, National Institute of
Standards and technology, Gaithersberg, USA, 1998.
Molkov, V. V., Grigorash, A. V., Eber, R. M., Tamanini, F., Dobashi, R. Vented gaseous
deflagrations with inertial vent covers: state o f the art and progress. Process Safety Progress, Vol.
23, No. 1. AIChemE, 2004.
Molkov, V. V., Grigorash, A. V., Eber, R. M., Tamanini, F., Dobashi, R. Vented gaseous
deflagrations: modelling o f translating inertial vent covers. J. Loss Prevention in the Process
Industries. Vol. 16, pp. 395 - 402. 2003.
Murdoch, P. Inspections: adding value to safety and maintenance. HazardEx, J. for Hazardous
Area Equipment. July, 2003.
418
Nowacki, J-E. Altemativa koldmedier - Safety with flammable refrigerants. Report no. 7. CIT
Chalmers Teknikpark, Goteborg, Sweden. 1995.
Orme, M., Liddament, M. W., Wilson, A. Numerical data for air infiltration and natural
ventilation calculations. Air Infiltration and Ventilation Centre, Coventry, UK, 1998.
PR 100374. Report on refrigerant release experiments. South Bank University7, London, UK,
October 2000.
Private communication regarding mobile air conditioner fire incident in the Philippines, 2002
(details withheld upon request).
Rabkinkov, V. A. The distribution o f flammable gas concentration in rooms. Fire Safety J., Vol.
13, pp. 211 -2 1 7 . 1988.
Radford, P. The Benefits of Refrigerant Inventory Control. Proc. Conf. Inst. Reffig., London.
1998.
Razus, D. M., Krause, U. Comparison of empirical and semi-empirical calculation methods for
venting gas explosions. Fire Safety7J., Vol. 36, No. 1. 2001.
Rew, P. J., Spencer, H. The sensitivity of risk assessment of flash fire events to modelling
assumptions. Proc. Hazards 14. Inst. Chem. Eng., London. November 1998.
Rew, P. J., Spencer, H., Daycock, J. Off-site ignition probability of flammable gases. J.
Hazardous Materials, Vol. 71, pp. 409 - 422. 2000.
Richard, R. Flammability hazard classification of refrigerants. Proc. IIR International Congress

Refrigeration, Washington, USA. 2003.
Ritter, T. J., Colboume, D. Quantitative Risk Assessment: Hydrocarbon Refrigerants. Proc. IIR
Gustav-Lorentzen Conf., Oslo, Norway. 1998.
Roberts, A. F. Methods for estimating harmful consequences o f fires and explosions following a
release of LPG. IR/F/FR/80/10. November 1980.
Sykes, R. I., Cerasoli, C. P., Henn, D. S. The representation of dynamic flow effects in a
Lagrangian puff dispersion model. J Hazardous Materials, Vol. 64, pp. 223 - 247. 1999.
Thomas, P. H. On strong two-dimensional plumes. Fire Safety Journal, Vol. 34, pp. 47-53. 2000.
Tickle, G. A. A model of the motion and dilution of a heavy gas cloud released on a uniform
slope in calm conditions. Journal of Hazardous Materials, Vol. 49, pp. 29-47, 1996.
419
UNEP. 2002 Report of the Technical Options Committee on Refrigeration, Air Conditioning and
Heat Pumps. United Nations Environment Programme, Nairobi, 2002.
Waterson, M. T. Marketing Pocketbook 2004. World Advertising Research Centre, Henley-on-

Thames, UK. 2003.
West Midlands Fire Service (WMFS), Fire Service Headquarters, Birmingham. Private
communication. July, 2003.
Wolfer, M., Seiler, H. Ammoniak und Kohlenwasserstoffe als kaltemittel: riskoanalyse,

produktehaftpflicht und straffecht. Bundesamtes fur Energie, Switzerland. December 1999.
Wolfer, M. Flammable working fluids in Switzerland: risk assessment of natural working fluids.
Proc. Workshop Proc. IEC Heat Pump Programme "Natural working fluids - a challenge for the
future”, Report no. HPC-WR-21, Paris, France, November 1999.
Woodward, J. L. Estimating the flammable mass of a vapour cloud. Centre for Chemical Process
Safety, American Institute of Chemical Engineers, New' York, USA, 1998.
Workplace Services. Inspectors' report o f the investigation of an accident at Cafe Grove, Comer
of Golden Grove Way and the Grove Way, Golden Grove, South Australia. South Australian
Department of Administrative and Information Services, January72000.
Zhao, Y., Shigang, Z., Guansan, T., Bin, L., Yitai, M. Study of refrigerant leakage in
refrigeration system. J. Fire Sciences, Vol. 20, pp. 237 - 245. May 2002.
420
APPENDICES
Appendix 1: Data from published QRAs
Table 55: Summary o f ORA results fo r stationary RAC equipment
Refer Frequency
Application Refrigerant Activity Consequence
ence (y 1) ’
operation fire on AHU/Cond U 3.2* 10'7
5?
ON
ON operation fire in cellar/garage 8.9X10*8
73 Split system operation fire in attic 8.2 * 10'9
6.5 kg
0> residential heat
l-i R32/R134a operation fire in closet 9.5X10-8
U pump
«S operation fire in room 9.0xl0'10
3
service fire 2.6 X10-6
operation fatality 8.3 xlO-8

r and Seiler
domestic heat
1 kg R290 service fatality 2.0 X10'7
pump
1999)
service injury 7.0X10-6

,<D
4-1
o operation fatality 8.5 xlO 8
£ indirect s/m system 20 kg R1270
operation damage 7.0x10'"
■s
<L> operation (7.5 m3 room) occupant fatality 1.8 x l0'7
o
ers and Verw
operation (50 m3 room) occupant fatality 8.0 xlO'8

(1997)
domestic heat
0.7 kg R290 operation (7.5 m3 room) public fatality 2.1X10-8
pump
service (7.5 m3 room) fatality 3.8x10'"
x>
w service (50 m3 room) fatality 3.0x10'
operation fatality within 2 m l.lx lO -6

“ 8
C operation fatality within 3 m 7.2x10'"
0) W"
^re *2
<u ice cream freezer 0.1 kg R290 operation fatality within 5 m 2.5 xlO'8
m |
c ,&£> operation fatality within 8 m 1.8 x l0'8
re
> >
service fatality 2.3 xlO'7
operation (50 m3 room) occupant fatality 6 .9xl0 '7

i?
S3 svs
ON operation (100 m3 room) occupant fatality 1.0 xlO'7
c
£(U *2
1> bulk milk tank 2.0 kg R290 operation (50 m3 room) public fatality 1.2 xlO'7
o 2
a C service (50 m3 room) fatality 9.0 xlO 7
re jo
>> service (100 m3 room) fatality 4.0X10‘7
421
Refer Frequency
Application Refrigerant Activity Consequence
ence (y 1) '
operation (20 m3 room) occupant fatality 2.3 xlO'7
operation (80 nr room) occupant fatality 1.0 xlO 7
domestic heat
1.5 kg R290 operation (20 m3 room) public fatality 2.6 x l0'8
pump
service (20 m3 room) fatality 5.0x10'"
service (80 m3 room) fatality 4.0xl0 '7
operation Fire and overpressure 1.3 xlO'5

Cold store 5.0 kg R290
service Fire and overpressure 1.0 xlO'9
Chest freezer 0.1 kg R290 operation Fire and overpressure 6.7x10'"

00
o Multideck freezer 0.4 kg R290 operation Fire and overpressure 8.4X10-6
operation Fire and overpressure 2.6 x l0'4

J Chiller cabinet 1.5 kg R290
9 service Fire and overpressure 4.2xl0'7
Remot draft chiller 0.2 kg R600a operation Fire and overpressure 6.6 xlO'9
Shelf draft chiller 0.1 kg R600a operation Fire and overpressure 9.6X10-6
Bottle cooler 0.3 kg R600a operation Fire and overpressure 4.8xl0'8
Table 56: Leak frequencies used in previous studies
Source Equipment details Leak frequency (y'1)

Domestic fridge during operation 2.9xl0'9
TR 14739 (2004)
Domestic fridge during transport 7.1xl0'6
Transport refrigeration, full bore hole 5.5xl0'5, 2.4X10-6
Jansen and van Gerw en Transport refrigeration. 10% full bore 1.8X10-4, 8.0x l0"5
(1996) Transport refrigeration, 1% full bore 5.5X10-4. 2.4xl0'4
Transport refrigeration, servicing l.OxlO'3
Heat pump full bore rupture 6.0x l0‘5
Elbers and Verwoerd (1997) Heat pump small leakage 6.0x l0'4
Heat pump servicing 3.8xl0'4
Supermarket chiller continuous 5.7xl0'5

Supermarket chiller spontaneous 6.3xl0'6
Wolfer and Seiler (1999)
Heat pump continuous 7.0xl0'4
Heat pump spontaneous 2.9xl0'4
Split heat pump, total 9.6xl0'3
Goetzler et al (1998) Split heat pump, total catastrophic 4.6xl0'4
Split heat pump, servicing l.OxlO'3
422
Cold stores, total 4.5xl0'3
Cold stores, servicing 1.4xl0'5
Supermarket chillers, total 2.0X10'1
Supermarket chillers, servicing 4.4xl0'3

ADL (1998) Chest freezer, total 9.3X10-6
Multi-deck freezer, total 1.2X10-4
Remote draft chiller, total 6.0X10-4
Shelf draft chiller, total 4.0x1 O’3
Bottle cooler, total 2.0x1 O’5
423
Appendix 2: Data for dispersion experiments
Table 57: Location o f sensors and corresponding representative volume ratio
Location (m) Ratio to define representative

Sampling volume for anemometer location
point
X V z (AL), Vt /V Rm
3.35 3.35 0 c6 AL] 0.0812

3.35 3.35 1 c7 a l2 0.0812
3.35 3.35 2 - a l3 0.0000
3.35 3.35 3 C8 AL., 0.0000
0.5 0.5 0 - a l5 0.0446
3.35 0.5 0 - a l6 0.0601
3.35 0.5 1 c9 a l7 0.1493
6.2 0.5 0 - ALg 0.0446
6.2 0.5 1 - ALg 0.0446
6.2 0.5 3 - ALio 0.0000
6.2 3.35 0 c5 AL„ 0.0601
6.2 3.35 1 - AL]2 0.0601
6.2 3.35 2 - a l ,3 0.0000
6.2 3.35 3 - AL m 0.0000
6.2 6.2 0 c3 AL]5 0.0446
6.2 6.2 1 - AL)6 0.0446
6.2 6.2 3 - a l 17 0.0000
3.35 6.2 0 C2 ALig 0.0601
3.35 6.2 1 c4 AL,9 0.0601
0.5 6.2 0 Cl AL20 0.0446
0.5 3.35 0 Cio AL21 0.0601
0.5 3.35 1 - AL22 0.0601
Figure 186: Room orientation and reference directions
424
Table 58: Parameter values and principle results for C 0 2 release tests
Unit air Mean

Unit Unit Fan Release Mass, r Rm Mass, r
Test flow Airflow Fan run room Release Release fm a x ^ fm a x
Unit type posi base disch. height cylinder room
no direction time (s) velocity position time (s)
tion (m) (m3 s'1) (m) (m) (kg) (kg m'3) (kg) (kg m'3) (kg m'3)
(m s'1)
9 none n/a 1.95 0 n/a 0 1.95 0.048 beneath 1.95 240 1.00 0.008 1.07 0.046 0.041
10 AHU#1 Wall 2.05 0.26 horizontal 210 2.35 0.198 beneath 1.95 210 1.40 0.007 0.94 0.018 0.009
13 AHU#1 Wall 2.05 0.26 horizontal 105 2,35 0.267 beneath 1.95 210 1.36 0.007 0.98 0.038 0.033
14 AHU#1 Wall 2.05 0.26 downwards 210 2.35 0.228 beneath 1.95 210 1.18 0.008 1.07 0.012 0.008
15 AHU#1 Wall 2.05 0.26 upwards 210 2.35 0.226 beneath 1.95 210 1.18 0.008 1.03 0.026 0.016
18 AHU#1 Wall 2.05 0.26 horizontal 210 2.35 0.240 beside end 1.95 210 1.00 0.008 1.05 0.028 0.013
21 AHU#1 Wall 0.50 0.26 horizontal 210 0,80 0.248 beneath 0.40 210 1.18 0.008 1.14 0.024 0.012
25 AHU#1 wall 2.05 0.26 horizontal 210 2.35 0.215 beneath 1.95 870 1.03 0.007 0.93 0.030 0.027
425
Unit air Mean
Unit Unit Fan Release Mass, Mass, c . r
Test flow Airflow Fan run disch. room Release Release cylinder C-Rm J.max J.max
Unit type posi base height room
no direction time (s) velocity position time (s)
tion (m) (m3 s'1) (m) (m) (kg) (kg in'3) (kg) (kg m'3) (kg m'3)
(m s'1)
31 AHU#1 wall 2.05 0.26 horizontal 210 2.35 0.200 room corner 1.95 210 1.03 0.010 1.30 0.046 0.030
35 AHU#3 ceiling 2.50 0.38 horizontal 210 2.50 0.263 beneath 2.20 210 0.60 0.005 0.62 0.007 0.005
36 AHU#3 ceiling 2.50 0.38 horizontal 210 2.50 0.281 beneath 2.20 210 2.02 0.013 1.72 0.017 0.015
37 AHU#3 ceiling 2.50 0.38 horizontal 210 2.50 0.234 wall 0.40 210 1.04 - - - -
38 Cabint#l floor 0.30 0.35 horizontal 210 0.30 0.254 beside end 0.20 210 1.24 0.005 0.72 0.020 0.008
41 Cabint#2 floor 2.00 0.05 horizontal 210 2.00 0.048 in cabinet (1.4) 210 0.80 0.007 0.97 0.095 0.084
42 Cabint#2 floor 2.00 0.05 horizontal 210 2.00 0.047 cond unit 2.10 210 0.70 0.005 0.67 0.018 0.012
43 none n/a 1.95 0 n/a 0 1.95 0.066 mid room 1.95 210 1.10 0.008 1.08 0.056 0.043
426
Appendix 3: Validation data from other sources
Notes for the Tables:

- Maximum concentrations were generally converted to C f max using equation (315).
- Equipment abbreviations: “PAC” - portable air conditioner, “SAC” - split air conditioner, “Dehum” - dehumidifier
- Position abbreviations: “mid” - middle/centre of room, “comer” - a comer of the room, “below” - beneath unit, “beside” - beside unit, “wall” -
against wall, “nr wall” - closer to wall than to the centre.
- No information was provided on unit position or unit base height in Clodic et al
- The first calculated C fmax was for a catastrophic release (tr cat = 210 s) and the second calculated Cfmax was the conversion to the specified leak
time.
- Room volume for Jabbour et al was 26 m3.
- The data from Kataoka et al (c) was from CFD calculations, but was considered acceptable forthis exercise.
Table 59: Test variables and calculated results fo r Aloisi

[a - Aloisi, September 1999, b - Aloisi, February 2000, c - Aloisi, May 1999]
Unit Unit air Air Fan Fan Release Room

Unit Release Release Mass Calc CfiHWX Calc C f max
Test no Gas type base flow outlet run disch. height volume f, max
type position time (s) (kg) tr (kg m'3)
(m) (m3 s'1) (m2) time (s) (m) (m) (m3) (kg m'3) ^r,cat ( k 8 m ' 3)
Ka) PAC R290 0.10 0.039 0.050 180 0.38 nr wall, below 0.5 240 0.35 36.3 0.0110 0.0082 0.0108
2(a) PAC R290 0.50 0.039 0.050 180 0.78 nr wall, below 0.5 240 0.35 36.3 0.0106 0.0262 0.0328
3(a) Dehum R290 0.10 0.042 0.058 180 0.28 nr wall, below 0.5 240 0.35 36.3 0.0153 0.0072 0.0096
4(a) Dehum R290 0.50 0.042 0.058 180 0.78 nr wall, below 0.5 240 0.35 36.3 0.0209 0.0261 0.0327
427
Unit Unit air Air Fan Fan Release Room Calc Cfmax
Unit Release Release Mass Calc C f max
Test no Gas type base flow outlet run discli. height volume ^ fm a x
type position time (s) (kg)
(m) (m3 s'1) (m2) time (s) (m) (m) (m3) (kg m'3) h ,c a t (kg m'3) tr (kg m'3)
5(a) SAC R290 1.50 0.042 0.048 180 1.5 nr wall, below 0.5 240 0.35 36.3 0.0223 0.0444 0.0547
1(b) Dehum R290 0.05 0.036 0.058 180 0.38 nr wall, beside 0.5 194 0.35 36.3 0.0248 - 0.0157
1(c) Dehum R290 0.05 0.010 0.058 0 0.38 nr wall, beside 0.5 282 0.23 36.3 0.0313 0.0441 0.0367
2(c) Dehum R290 0.05 0.056 0.058 282 0.38 nr wall, beside 0.5 282 0.23 36.3 0.0067 0.0084 0.0081
3(c) PAC R290 0.05 0.042 0.058 282 0.38 nr wall, beside 0.5 282 0.23 36.3 0.0033 0.0111 0.0103
4(c) PAC R290 0.05 0.042 0.058 180 0.38 nr wall, beside 0.5 282 0.23 36.3 0.0076 0.0111 0.0103
Table 60: Test variables and calculated results fo r Clodic et al

[a - Clodic and Cai, 1997, b - Clodic and Cai, 1996, c - Clodic, 1997]
Unit air Air Fan Release Room r

Gas Fan run Release Release Mass fm a x Calc C j?max Calc C- f max
Test no Unit type flow outlet disch. height volume
type time (s) position time (s) (kg) (kg in’3) tr (kg in'3)
(m3 s'1) (m2) (in) (m) (m3) (kg m'3) t r ,ca,
1(a) n/k R290 0.00 n/k 0 0.05 mid 0.05 384 0.13 16 0.0260 0.8346 0.4842
2(a) n/k R290 0.00 n/k 0 0.05 mid 0.05 768 0.13 16 0.0260 0.8346 0.2591
3(a) n/k R290 0.00 n/k 0 0.05 mid 0.05 192 0.13 16 0.0329 0.8346 0.9049
4(a) n/k R290 0.00 n/k 0 0.05 mid 0.05 1536 0.13 16 0.0370 0.8346 0.1386
21(a) n/k R600a 0.00 n/k 0 0.05 mid 0.05 1536 0.13 16 0.0488 0.8346 0.1386
27 (a) n/k R600a 0.00 n/k 0 0.05 mid 0.05 768 0.13 16 0.0361 0.8346 0.2591
428
Unit air Air Fan Release Room
Gas Fan run Release Release Mass ^ fmax Calc C fmm Calc Cfmax
Test no Unit type flow outlet discli. height volume
type time (s) position time (s) (kg) tr (kg m'3)
(m3 s'1) (m2) (m) (m) (m3) (kg m'3) tr.cat (k8 m’3)
9(a) n/k R600a 0.00 n/k 0 0.05 mid 0.05 384 0.13 16 0.0542 0.8346 0.4842
10 (a) n/k R600a 0.00 n/k 0 0.05 mid 0.05 192 0.13 16 0.0542 0.8346 0.9049
3(a) n/k R290 0.00 n/k 0 0.05 mid 0.05 192 0.13 16 0.0329 0.8346 0.9049
7(a) n/k R290 0.00 n/k 0 0.30 mid 0.30 192 0.13 16 0.0233 0.1391 0.1489
5 (a) n/k R290 0.00 n/k 0 0.70 mid 0.70 192 0.13 16 0.0110 0.0596 0.0631
6(a) n/k R290 0.00 n/k 0 1.50 mid 1.50 192 0.13 16 0.0082 0.0278 0.0290
12 (a) n/k R290 0.00 n/k 0 0.05 corner 0.05 384 0.13 16 0.0576 0.8346 0.4842
13(a) n/k R600a 0.00 n/k 0 0.05 corner 0.05 384 0.13 16 0.0813 0.8346 0.4842
14 (a) n/k R290 0.00 n/k 0 0.05 corner 0.05 384 0.13 16 0.0452 0.8346 0.4842
2(b) PAC R290 0.00 0.04 0 0.28 corner 0.28 480 0.24 60 0.0466 0.1303 0.0659
3(b) PAC R290 0.00 0.04 0 0.28 corner 0.28 360 0.24 60 0.0480 0.1303 0.0835
4(b) PAC R290 0.00 0.04 0 0.28 corner 0.28 576 0.24 60 0.0548 0.1303 0.0567
6(b) PAC R290 0.01 0.04 360 0.28 corner 0.28 360 0.24 60 0.0055 0.0339 0.0238
7(b) PAC R290 0.01 0.04 360 0.28 corner 0.28 360 0.24 60 0.0548 0.0339 0.0238
8(b) PAC R290 0.00 0.04 0 0.28 corner 0.28 360 0.24 60 0.0548 0.1303 0.0835
9(b) PAC R290 0.01 0.04 576 0.28 corner 0.28 576 0.24 60 0.0192 0.0339 0.0175
15 (b) PAC R290 0.01 0.04 360 0.28 corner 0.28 360 0.24 60 0.0329 0.0339 0.0238
16(b) PAC R290 0.00 0.04 0 0.28 corner 0.28 1440 0.24 60 0.0302 0.1303 0.0266
17(b) PAC R290 0.01 0.04 360 0.28 corner 0.28 360 0.24 60 0.0822 0.0218 0.0160
18(b) PAC R290 0.00 0.04 0 0.28 corner 0.28 720 0.24 60 0.0411 0.1303 0.0472
19(b) PAC R290 0.01 0.04 720 0.28 corner 0.28 720 0.24 60 0.0411 0.0339 0.0151
429
Unit air Air Fan Release Room
Gas Fan run Release Release Mass C fm a x CalcC fmax Calc Cf max
Test no Unit type flow outlet disch. height volume
type time (s) position time (s) (kg) trcat (kgin3) tr (kg m'3)
(m3 s'1) (m2) (m) (m) (m3) (kg m'3)
20 (b) PAC R290 0.00 0.04 0 0.28 corner 0.28 360 0.24 60 0.0822 0.1303 0.0835
8(c) n/k R290 0.00 n/k 0 0.05 mid 0.05 192 0.13 16 0.0110 0.0246 0.0256
11(c) n/k R290 0.00 n/k 0 0.05 mid 0.05 192 0.13 16 0.0247 0.8346 0.9049
Table 61: Test variables and calculated results fo r Jabbour et al, 2003
Air c fm a x Calc C j-max Calc Cfmax

Test Airflow Air flow Fan run Fan disch. Release Release Release
Gas type outlet Mass (kg)
no source (m3 s'1) time (s) (m) position height (m) time (s) (kg m tr (kg m'3)
(m2) (kg in'3) tr,cat ' 3)
39 R290 0 0.00 0.037 0 0.68 mid 0.54 258 0.150 0.0781 0.0627 0.0535
40 R290 0 0.00 0.037 0 0.68 mid 0.54 231 0.150 0.0776 0.0627 0.0583
41 R290 0 0.00 0.037 0 0.68 mid 0.54 280 0.338 0.0918 0.1412 0.1133
42 R290 0 0.00 0.048 0 0.60 wall 0.60 235 0.150 0.0521 0.0568 0.0522
43 R290 0 0.00 0.048 0 0.60 wall 0.60 247 0.150 0.0480 0.0568 0.0503
61 R290 0 0.05 0.037 564 0.68 mid 0.54 564 0.410 0.0137 0.0292 0.0191
62 R290 0 0.00 0.037 0 0.68 mid 0.54 740 0.410 0.0356 0.1713 0.0654
63 R290 0 0.05 0.048 692 0.60 wall 0.60 692 0.410 0.0123 0.0242 0.0155
64 R290 0 0.00 0.048 0 0.60 wall 0.60 513 0.410 0.0507 0.1553 0.0793
65 R290 0 0.05 0.048 1015 1.00 wall 1.00 1015 0.410 0.0082 0.0199 0.0123
66 R290 0 0.00 0.048 0 1.00 wall 1.00 1031 0.410 0.0356 0.0932 0.0316
68 R290 0 0.00 0.048 0 1.80 wall 1.80 548 0.410 0.0493 0.0518 0.0299
69 R290 0 0.00 0.048 0 1.80 wall 1.80 590 0.410 0.0480 0.0518 0.0287
1 R600a 0 0.00 0.037 0 0.68 mid 0.54 234 0.118 0.0578 0.0493 0.0454
430
Air Calc Cfmax Calc C f max
Test Airflow Air flow Fan run Fan disch. Release Release C fm a x
no source (m3 s'1) time (s) (m) height (m) time (s) (kg in3)
(m2) (kg m'3) fr.cat 0 ^ 8 m ' 3) tr
3 R600a 0 0.00 0.037 0 0.68 0.54 239 0.150 0.0777 0.0627 0.0568
4 R600a 0 0.00 0.037 0 0.68 0.54 308 0.150 0.0560 0.0627 0.0468
5 R600a 0 0.00 0.037 0 0.68 0.54 169 0.150 0.0596 0.0627 0.0740
7 R600a 0 0.00 0.037 0 0.68 0.54 256 0.383 0.1174 0.1600 0.1375
8 R600a 0 0.00 0.037 0 0.68 0.54 406 0.383 0.0993 0.1600 0.0967
106 R600a 0 0.00 0.037 0 0.68 0.54 224 0.172 0.0939 0.0718 0.0684
107 R600a 0 0.00 0.037 0 0.68 0.54 232 0.172 0.1011 0.0718 0.0666
108 R600a 0 0.00 0.037 0 0.68 0.54 208 0.172 0.1084 0.0718 0.0724
109 R600a 0 0,00 0.037 0 0.60 0.50 111 0.086 0.0704 0.0391 0.0640
110 R600a 0 0.00 0.037 0 0.60 0.50 54 0.043 0.0524 0.0195 0.0559
9 R600a 0 0.00 0.048 0 0.60 0.60 248 0.150 0.0470 0.0568 0.0501
10 R600a 0 0.00 0.048 0 0.60 0.60 151 0.150 0.0524 0.0568 0.0728
111 R600a 0 0.00 0.048 0 0.60 0.60 222 0.172 0.0506 0.0651 0.0625
112 R600a 0 0.00 0.048 0 0.60 0.60 254 0.172 0.0524 0.0651 0.0565
113 R600a 0 0.00 0.048 0 0.60 0.60 264 0.383 0.0524 0.1451 0.1221
114 R600a 0 0.05 0.048 275 0.60 0.60 275 0.383 0.0397 0.0312 0.0275
115 R600a 0 0.00 0.048 0 0.60 0.60 245 0.383 0.0524 0.1451 0.1292
116 R600a 0 0.05 0.048 294 2.20 2.20 294 0.383 0.0235 0.0376 0.0316
117 R600a 0 0.00 0.048 0 2.20 2.20 227 0.383 0.0325 0.0396 0.0380
11 R600a 0 0.05 0.048 222 0.60 0.60 222 0.383 0.0181 0.0312 0.0304
12 R600a 0 0.00 0.048 0 0.60 0.60 315 0.383 0.0759 0.1451 0.1069
431
Air r Calc Cfmax Calc Cfmax
Test Airflow Air flow Fan run Fan disch. Release Release Release fm a x
no source (m3 s'1) time (s) (m) position height (m) time (s) (kg nr3) tr (kg m'3)
(m2) (kg m'3) tr, cut
51 R290 mech vent 0.04 0.031 232 1.77 wall 0.60 232 0.300 0.0548 0.0374 0.0353
52 R290 mech vent 0.08 0.031 267 1.77 wall 0.60 267 0.450 0.0589 0.0326 0.0294
53 R290 mech vent 0.08 0.031 267 1.77 wall 1.00 267 0.550 0.0274 0.0205 0.0195
76 R290 mech vent 0.04 0.031 422 2.43 mid 0.60 422 0.280 0.0960 0.0652 0.0405
75 R290 mech vent 0.11 0.031 453 2.43 mid 0.60 453 0.605 0.0973 0.0510 0.0354
74 R290 mech vent 0.21 0.031 922 2.43 mid 0.60 922 1.050 0.0754 0.0482 0.0315
82 R290 mech vent 0.11 0.031 482 2.43 wall 0.60 482 0.605 0.0699 0.0510 0.0344
81 R290 mech vent 0.21 0.031 675 2.43 wall 0.60 675 1.050 0.0493 0.0482 0.0344
87 R290 mech vent 0.02 0.031 308 2.43 wall 1.80 308 0.403 0,0384 0.0467 0.0378
88 R290 mech vent 0.02 0.031 543 2.43 wall 1.80 543 0.410 0.0452 0.0448 0.0268
90 R290 mech vent 0.04 0.031 685 2.43 wall 1.80 685 0.490 0.0096 0.0311 0.0195
86 R290 mech vent 0.09 0.031 724 2.43 wall 1.80 724 0.700 0.0192 0.0213 0.0182
93 R290 mech vent 0,04 0.031 760 2.43 wall 2.20 760 0.555 0.0137 0.0284 0.0187
94 R290 mech vent 0.09 0.031 652 2.43 wall 2.20 652 0.770 0.0137 0.0186 0.0182
47 R290 mech vent 0.04 0.031 264 1.77 mid 0.60 264 0.300 0.0740 0.0374 0.0328
48 R290 mech vent 0.08 0.031 242 1.77 mid 0.60 242 0.450 0.0658 0.0326 0.0307
49 R290 mech vent 0,04 0.031 238 1.77 wall 0.60 238 0.300 0.0493 0.0374 0.0348
50 R290 mech vent 0.08 0.031 257 1.77 wall 0.60 257 0.450 0.0631 0.0326 0.0299
34 R600a mech vent 0.04 0.031 269 1.77 wall 0.60 269 0.300 0.0596 0.0635 0.0538
35 R600a mech vent 0.06 0.031 427 1.77 wall 0.60 427 0.450 0.0542 0.0534 0.0359
36 R600a mech vent 0.02 0.031 307 1.77 wall 1.00 307 0.300 0.0451 0.0614 0.0477
432
Air rV CalcC fmax Calc C f max
Test Airflow Air flow Fan run Fan disch. Release Release fm a x
no source (m3 s'1) time (s) (m) height (m) time (s) tr (kg in"3)
(m2) (kg m'3) t r ,cat (kg m ' 3)
37 R600a mech vent 0.02 0.031 244 1.77 1.00 244 0.300 0.0488 0.0614 0.0556
38 R600a mech vent 0.02 0.031 469 1.77 1.80 469 0.450 0.0343 0.0505 0.0326
24 R600a mech vent 0.04 0.031 220 1.77 0.60 220 0.300 0.0596 0.0635 0.0615
25 R600a mech vent 0.06 0.031 440 1.77 0.60 440 0.450 0.0686 0.0534 0.0353
26 R600a mech vent 0.04 0.031 222 1.77 0.60 222 0.300 0.0668 0.0635 0.0611
27 R600a mech vent 0.06 0.031 535 1.77 0.60 535 0.450 0.0415 0.0534 0.0317
28 R600a mech vent 0.06 0.031 405 1.77 0.60 405 0.450 0.0506 0.0534 0.0370
29 R600a mech vent 0.06 0.031 258 1.77 0.60 258 0.450 0.0668 0.0534 0.0476
31 R600a mech vent 0.02 0.031 318 1.77 1.00 318 0.300 0.0415 0.0614 0.0466
30 R600a mech vent 0.05 0.031 442 1.77 1.00 442 0.450 0.0415 0.0388 0.0271
32 R600a mech vent 0.02 0.031 394 1.77 1.80 394 0.450 0.0343 0.0505 0.0359
33 R600a mech vent 0.05 0.031 425 1.77 1.80 425 0.600 0.0451 0.0271 0.0222
17 R600a mech vent 0.04 0.031 233 2.43 0.60 233 0.300 0.1264 0.1061 0.0982
18 R600a mech vent 0.06 0.031 281 2.43 0.60 281 0.450 0.1047 0.0892 0.0737
19 R600a mech vent 0.04 0.031 193 2.43 0.60 193 0.300 0.0596 0.1061 0.1130
20 R600a mech vent 0.06 0.031 295 2.43 0.60 295 0.450 0.0668 0.0892 0.0714
21 R600a mech vent 0.01 0.031 258 2.43 1.80 258 0.400 0.0325 0.0505 0.0449
22 R600a mech vent 0.02 0.031 296 2.43 1.80 296 0.450 0.0361 0.0568 0.0467
23 R600a mech vent 0.05 0.031 441 2.43 1.80 441 0.600 0.0542 0.0452 0.0325
70 R290 nat vent 0.00 0.025 0 0.60 0.60 140 0.115 0.0877 0.0436 0.0591
44 R290 nat vent 0.00 0.025 0 0.60 0.60 215 0.150 0.0685 0.0568 0.0558
433
Air r Calc C J]max Calc C fmax
Test Airflow Air flow Fan run Fan disch. Release Release Release fm a x
no source (m3 s'1) time (s) (m) position height (m) time (s) tr (kg m'3)
(m2) (kg m 3) ^r ,cat (kg m ' 3)
71 R290 nat vent 0.00 0.025 0 0.60 wall 0.60 132 0.115 0.0493 0.0436 0.0618
45 R290 nat vent 0.00 0.025 0 0.60 wall 0.60 232 0.150 0.0411 0.0568 0.0527
46 R290 nat vent 0.00 0.025 0 0.60 wall 0.60 204 0.150 0.0466 0.0568 0.0581
73 R290 nat vent 0.00 0.025 0 1.80 wall 1.80 361 0.327 0.0260 0.0413 0.0303
13 R600a nat vent 0.00 0.025 0 0.60 mid 0.60 207 0.150 0.0921 0.0568 0.0574
14 R600a nat vent 0.00 0.050 0 0.60 mid 0.60 252 0.150 0.0759 0.0568 0.0495
15 R600a nat vent 0.00 0.025 0 0.60 wall 0.60 208 0.150 0.0379 0.0568 0.0572
16 R600a nat vent 0.00 0.025 0 1.00 wall 1.00 242 0.224 0.0433 0.0509 0.0462
72 R290 nat vent 0.00 0.025 0 0.60 wall 0.60 412 0.410 0.0548 0.1553 0.0935
79 R290 mech vent 0.04 0.031 583 2.43 mid 0.60 583 0.410 0.0699 0.0955 0.0475
80 R290 mech vent 0.05 0.048 580 0.60 mid 0.60 580 0.410 0.0137 0.0242 0.0166
85 R290 mech vent 0.04 0.031 611 2.43 wall 0.60 611 0.410 0.0617 0.0955 0.0460
84 R290 mech vent 0.05 0.048 592 0.60 wall 0.60 592 0.410 0.0137 0.0242 0.0165
89 R290 mech vent 0.04 0.031 682 2.43 wall 1.80 682 0.490 0.0096 0.0311 0.0196
Table 62: Test variables and calculated results fo r Kataoka et al

[a - Kataoka, 1999, b - Kataoka et al, 2000b, c - Kataoka, 2001]
Fan disch. Room c^ fm. a x Calc C fmax Calc C fmax

Air flow Fan run Release Release Release
Test no Gas type Mass (kg) volume
(m3 s'1) time (s) (m) position height (m) time (s) tr (kg in'3)
(m3) (kg m'3) tr.cat (kg m'3)
1(a) R290 0.000 0 1.80 n/k 1.80 240 0.22 18 0.0286 0.0419 0.0393
2(a) R290 0.000 240 0.60 n/k 0.60 240 0.07 18 0.0595 0.0421 0.0384
Room c /.max
. Calc Cfmax Calc Cfmax
Air flow Fan run Fan disch. Release Release Release
(m3 s 1) time (s) (m) position height (m) time (s) tr (kg m'3)
(m3) (kg m'3) tr.cat m ' 3)
3(a) R290 0.202 0 1.80 n/k 1.80 240 1.22 18 0.0589 0.0669 0.0669
4(a) R290 0.000 0 1.80 n/k 1.80 240 1.22 18 0.0781 0.2339 0.2198
5 (a) R290 0.000 240 1.80 wall 1.80 240 0.22 18 0.0284 0.0419 0.0393
6(a) R290 0.017 240 1.80 wall 1.80 240 0.25 18 0.0288 0.0480 0.0451
Kb) R134a 0.000 0 0.10 mid 0.10 602 0.62 17 0.7509 2.1472 0.8600
2(b) R134a 0.000 0 0.10 mid 0.10 1560 1.30 17 1.0430 4.4877 0.7858
3(b) R134a 0.000 0 0.50 mid 0.50 1560 1.30 17 0.3546 0.8975 0.2175
4(b) R134a 0.000 0 0.50 mid 0.50 1560 1.30 17 0.4380 0.8975 0.2175
5(b) R134a 0.000 0 0.50 mid 0.50 1560 1.30 17 0.4589 0.8975 0.2175
6(b) R134a 0.000 0 0.18 mid 0.18 1114 1.30 17 0.8344 2.4932 0.6302
7(b) R134a 0.000 0 0.90 mid 0.90 1114 1.30 17 0.2503 0.4986 0.1812
8(b) R134a 0.000 0 0.50 mid 0.50 1114 1.30 17 0.4380 0.8975 0.2759
9(b) R134a 0.000 0 0.10 mid 0.10 1114 1.30 17 1.0847 4.4877 1.0526
10(b) C02 0.000 180 0.10 mid 0.10 180 0.15 17 0.2886 0.5178 0.5920
11(b) C02 0.017 0 0.10 mid 0.10 180 0.15 17 0.0144 0.0112 0.0114
1(c) R290 0.000 0 0.50 mid 0.50 240 0.07 15 0.0776 0.0505 0.0460
2(c) R290 0.000 0 0.50 mid 0.50 240 0.07 15 0.0794 0.0505 0.0460
3(c) R290 0.000 0 0.50 mid 0.50 120 0.13 15 0.1371 0.0864 0.1274
3a (c) R290 0.000 0 0.50 mid 0.50 240 0.15 15 0.0000 0.1037 0.0945
4 (c) R290 0.000 0 0.50 mid 0.50 240 0.25 15 0.1714 0.1728 0.1576
5(c) R290 0.000 0 0.50 mid 0.50 240 0.23 154 0.0379 0.0498 0.0446
435
Room r fmax CalcC fmax Calc C f max
Air flow Fan run Fan disch. Release Release Release
(m3 s'1) time (s) (m) position height (m) time (s) tr,Cat (kgm'3) tr (kg m 3)
(m3) (kg m'3)
6(c) R290 0.000 0 1.80 mid 1.80 240 0.22 15 0.0379 0.0419 0.0396
7(c) R290 0.000 0 1.80 mid 1.80 240 0.68 154 0.0379 0.0415 0.0379
8(c) R290 0.000 0 1.80 mid 1.80 240 0.93 154 0.0451 0.0564 0.0515
9(c) R32 0.000 0 0.50 mid 0.50 240 1.09 15 0.6083 0.7536 0.6870
10(c) R32 0.000 0 0.50 mid 0.50 240 3.41 154 0.2680 0.7456 0.6676
11(c) R32 0.000 0 1.80 mid 1.80 240 3.26 15 0.3871 0.6261 0.5921
12 (c) R32 0.000 0 1.80 mid 1.80 240 3.26 15 0.3956 0.6261 0.5921
13(c) R32 0.000 0 1.80 mid 1.80 240 10.20 154 0.2829 0.6195 0.5662
14(c) R152a 0.000 0 0.50 mid 0.50 240 0.26 15 0.2052 0.1825 0.1664
15(c) R152a 0.000 0 0.50 mid 0.50 240 0.83 154 0.0945 0.1815 0.1625
16 (c) R152a 0.000 0 1.80 mid 1.80 240 0.88 15 0.1080 0.1686 0.1595
17(c) R152a 0.000 0 1.80 mid 1.80 240 2.49 154 0.0918 0.1512 0.1382
18(c) R600a 0.000 0 0.50 mid 0.50 240 0.07 15 0.0737 0.0505 0.0460
19(c) R600a 0.000 0 0.50 mid 0.50 240 0.23 154 0.0356 0,0501 0.0448
20 (c) R600a 0.000 0 1.80 mid 1.80 240 0.22 15 0.0356 0.0421 0.0398
21(c) R600a 0.000 240 1.80 mid 1.80 240 0.69 154 0.0333 0.0417 0.0381
22 (c) R290 0.000 240 1.80 mid 1.80 240 1.22 15 0.0974 0.2339 0.2212
23 (c) R290 0.017 240 1.80 mid 1.80 240 1.22 15 0.0776 0.2339 0.2212
24 (c) R290 0.017 240 1.80 mid 1.80 240 0.35 15 0.0487 0.0663 0.0627
25 (c) R290 0.000 240 1.80 mid 1.80 240 1.18 154 0.0577 0.0718 0.0657
26 (c) R290 0.171 240 1.80 mid 1.80 240 1.18 154 0.0505 0.0133 0.0129
436
Room r fmax Calc C j'max Calc C f max
Air flow Fan ran Fan disch. Release Release Release
(m3 s'1) time (s) (m) position height (m) time (s) tr,cat (k8 m 3) tr (kg m'3)
(m3) (kg in 3)
27 (c) R32 0.000 240 1.80 mid 1.80 240 15.20 154 0.3722 0.9231 0.8438
28 (c) R32 0.171 0 1.80 mid 1.80 240 15.20 154 0.3530 0.2079 0.1994
437
Appendix 4: Properties of flammable refrigerants
Table 63 provides flammable properties of various substances used within this Part 3. The data
for HCs was from Woodward (1998) and from Richard (2003) for fluorocarbons. The density'
constant was obtained from density' data generated using Refprop (McLinden et al, 1998).
Table 63: Flammable properties o f hydrocarbons and selected refrigerants
Methan Difluoro
Iso Difluoro
Chemical name Propane Propene Hexane
butane e -ethane
methane
Refrigerant number R600a R290 R1270 R50 - R32 R152a
Mol mas (kgkmol1) 58.1 44.1 42.1 16.0 86.2 52.0 66.1
LFL (kg m"3) 0.045 0.038 0.043 0.032 0.039 0.27 0.121
UFL (kg m'3) 0.203 0.172 0.173 0.097 0.277 0.52 0.436
C stoic (kg m'3) 0.074 0.072 0.077 0.062 0.076 0.37 0.208
H (kJkg1) 45580 46320 45805 50020 44700 9400 16900
H ,mo1 (kJkmol1) 134600 103900 20400 74900 167300 - -
Tfiam (K) 2400 2394 2506 2225 2400 2300 2325
U flan,Jar, 0 " s ’) 0.37 0.43 0.48 0.35 0.44 0.06 0.23
Density constant a 0.515 0.354 0.322 -0.802 0.629 0.450 0.571
The density' of a refrigerant and (dry ) air mixture at room temperature and standard atmospheric
pressure is (equation 408)
p = \ 2 + a -C (408)
where C is the refrigerant concentration (kg m°)
For flammable mixtures at concentrations other than stoichiometric, properties were estimated
using the methods below.
Heat of combustion
The heat of combustion was calculated with equation (409), w'here only the quantity' of
refrigerant or oxy'gen available for stoichiometric combustion (depending upon which w as in
excess) was included, i.e., the energy from excess fuel or oxy'gen was discounted.
’ mol X 1 a 7 mo/ mol

H = t t
' “ \,i 2 -i 2j
t j t
2>'
(409)
M- N™1
438
where M is the molar mass, N mo1 is the number of moles, H 'mot is the molar enthalpy of
formation. Subscript 1 represents unbumed species and subscript 2 the combustion products.
(For the combustion products, carbon dioxide H 'mo1 = 393800 kJ km ol1 and water vapour
H ,m°i = 242000 kJkm ol1.)
Adiabatic flame temperature

The adiabatic flame temperature was found from the following energy balance (equation 410).
(410)
where X ] is the mass fraction of the reactants, and X 2 the mass fraction of the combustion
products, AHl is the enthalpy of reactants (at initial air temperature), H x is the heat of
combustion, and Cp2 is the specific heat of combustion products.
Laminar flame velocity

Flame velocity is known to be a function of the density , thermal conductivity' and specific heat of
the initial mixture, as well as the combustion rate. Across the flammable range, these properties
van' little, so ^ is adjusted according to the combustion rate only. Since the combustion
rate is related to expansion ratio, and therefore Tflam, uâmlam is approximated from the variation
m Tflam (equation 411).
11 flam,lam ( O ~ flam P st ’ O ' (411)

v ^ T flam P s t ■>C pi ) ^
where AT ^ refers to the difference betw een the stoichiometric T flam and the value at the
reference concentration, or the concentration of the flammable limit ( ) adjacent to the
reference concentration.
439
Appendix 5: Data for overpressure calculations
Table 64: Data from overpressure measurements and model results
mF VF v
YRni -fcv, A wvp Pvp,fix Ap°meas
Ref No Gas &P°calc
(m3)
(VF)U‘ (m2)
(kg) (m3) (kgm2) (kPa) (kPa) (kPa)
1 hex 0.136 0.2 20.7 1.0 1 2.50 0 5.2 5.5

2 hex 0.139 0.2 20.7 1.0 1 2.50 0 5.9 5.7
De
3 hex 0.13 0.2 20.7 1.0 1 2.50 0 4.7 4.7
Haan
4 hex 0.158 0.3 20.7 1.0 1 2.50 0 4.7 5.0
5 hex 0.153 0.3 20.7 1.0 1 2.50 0 5.8 4.8
1 R50 0.06 1.0 1.0 0.5 0.16 0.00 50 140 98.1
2 R50 0.06 1.0 1.0 0.5 0.36 0.00 50 64 51.3
3 R50 0.06 1.0 1.0 0.5 0.10 0.00 10 140 153.2
4 R50 0.06 1.0 1.0 0.5 0.10 0.00 20 219 157.8
5 R50 0.06 1.0 1.0 0.5 0.10 0.00 50 195 172.4
6 R50 0.06 1.0 1.0 0.5 0.20 0.00 10 84 57.6
7 R50 0.06 1.0 1.0 0.5 0.20 0.00 20 148 60.9
Razus
8 R50 0.06 1.0 1.0 0.5 0.20 0.00 50 111 74.1
#1
9 R50 0.06 1.0 1.0 0.5 0.30 0.00 10 43 29.4
10 R50 0.06 1.0 1.0 0.5 0.30 0.00 20 77 31.7
11 R50 0.06 1.0 1.0 0.5 0.30 0.00 50 59 51.3
12 R50 0.06 1.0 1.0 0.5 0.05 0.00 32 200 317.0
13 R50 0.06 1.0 1.0 0.5 0.10 0.00 16 100 149.4
14 R50 3.05 49.1 49.1 0.8 3.46 0.00 0 12 17.8
15 R50 2.08 33.5 33.5 0.5 2.57 0.00 0 15 39.3
1 R50 0.16 2.4 2.4 0.5 0.26 0.00 14 6 41.1
2 R50 0.15 2.3 2.3 1.0 0.37 0.00 0 6 5.4
3 R50 0.15 2.3 2.3 1.0 0.16 0.00 0 21 26.4
4 R50 0.15 2.3 2.3 1.0 0.09 0.00 0 44 67.2
Razus 5 R50 0.04 0.7 0.7 1.0 0.19 0.00 0 3 4.1

#2 6 R50 0.04 0.7 0.7 1.0 0.08 0.00 0 15 23.0
7 R50 0.04 0.7 0.7 1.0 0.04 0.00 0 35 64.1
8 R50 2.01 30.4 30.4 0.5 1.33 0.00 1 21 42.5
9 R50 2.03 30.4 30.4 1.0 2.74 0.00 2 22 2.8
10 R50 2.05 30.4 30.4 1.0 1.33 0.00 4 54 11.7
Razus 1 R290 0.07 1.0 1.0 0.5 0.20 0.00 10 100 96.5
n
2 R290 0.07 1.0 1.0 0.5 0.40 0.00 10 32 32.5
3 R290 0.07 1.0 1.0 0.5 0.60 0.00 10 14 15.4
4 R290 14.70 203.8 203.8 1.0 21.58 0.00 0 3 3.5
440
mF V,Rm 4
wvp Pvp,fix
No Gas
(m3) 3\
(nr) 0' ' ) (m2)
(kg) (kgm'2) (kPa) (kPa) (kPa)
5 R290 14.70 203.8 203.8 1.0 17.30 0.00 0 6 5.5

6 R290 0.80 11.0 11.0 1.0 1.36 0.00 5 9 17.6
7 R290 2.44 30.4 30.4 0.5 0.58 0.00 40 70 379.5
8 R290 0.06 0.7 0.7 0.5 0.02 0.00 65 522 596.6
9 R290 0.06 0.7 0.7 0.5 0.02 0.00 65 531 596.6
10 R290 0.06 0.7 0.7 0.5 0.03 0.00 72 559 478.2
11 R290 0.06 0.7 0.7 0.5 0.07 0.00 86 314 255.7
12 R290 0.06 0.7 0.7 0.5 0.10 0.00 86 275 185.7
13 R290 0.00 0.0 0.0 0.5 0.03 0.00 10 19 32.1
14 R290 0.07 0.8 0.8 0.5 0.29 0.00 0 5 32.7
15 R290 3.31 35.0 35.0 1.0 1.00 0.00 0 75 83.4
16 R290 3.31 35.0 35.0 0.5 1.00 0.00 0 137 209.4
441

33927175

Uploaded by

Copyright:

Available Formats

33927175

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

33927175

Uploaded by

Copyright:

Available Formats

19 1727310 5

UNIVERSITY OF LONDON THESIS

Degree P ^ D > Year T }jD O 6 Name of Author }

C. 1975 - 1988. Most theses may be copied upon completion of a Copyright

D. 1989 onwards. Most theses may be copied.

This thesis comes within category D.

□ This copy has been deposited in the Library of

Daniel Henry Colbourne

All rights reserved

INFO RM ATIO N TO ALL USERS

Daniel Henry Colbourne.

Date: 20thJune 2006

PART 1: METHODS FOR EVALUATING THE GLOBAL WARMING IMPACT OF

PART 2: DEVELOPING A NEW SYSTEM BALANCING TECHNIQUE FOR MIXTURE

PART 3: APPRAISING THE FLAMMABILITY HAZARDS OF HYDROCARBON

Table 1: GHG (refrigerant) data from IPCC/TEAP (2005) 36

Figure 1: Identification o f sources and types o f GHG emissions 42

Motivation and description of work

Extending Methods fo r Evaluating the Global Warming Impact o f Refrigerating Systems.

Developing a New System Balancing M ethod fo r M ixture Refrigerants.

Appraising the Flammability Hazards o f Hydrocarbon Refrigerants.

area for two-phase heat transfer (m2)

b constant rating (W)

B app application basic rating (W K '1)

Cp specific heat at constant pressure (kJ kg'1 K '1)

E comp compressor energy consumption (kW h)

F radiative forcing (W m'2 kg'1)

specific enthalpy at evaporator inlet (J kg"1)

k thermal conductivity (W m '1 K '1)

B trav travel distance (km)

n index, sensitivity factor

Non, number of components (-)

N xn. number of service/maintenance visits (-)

N start number of compressor starts (-)

N tpup number of refrigerant top-ups (-)

Ap pressure drop (Pa)

Pdisch pressure at compressor discharge (Pa)

Psuet pressure at compressor suction (Pa)

Pvent probability that refrigerant is vented (-)

O heat flow (W)

Q<m application heat load (W)

Qc condensing capacity (W)

Qc,comp compressor condensing capacity (W)

Qcomp compressor capacity (W)

Qe evaporating capacity (W)

Qc,comp compressor evaporating capacity (W)

AR mean variation in leak ratio (-)

Rdist leak mass fraction during refrigerant distribution (-)

Re Reynolds number (-)

Sou solubility of refrigerant in oil (kg kg'1)

Sw wall shear stress (N m'2)

TOPP application temperature (°C)

Tdisch temperature at compressor discharge (°C)

TLT1 total lifetime temperature impact (K)

^ a fc off-cvcle time (s)

& onc on-cycle time (s)

*op operating time for compressor wear (s)

ATm thermostat differential temperature setting (K)

compressor start-up time (s)

T~* temperature at compressor suction (°C)