Professional Documents
Culture Documents
IP RESEARCH REPORT
IGNITION PROBABILITY REVIEW, MODEL DEVELOPMENT AND
LOOK-UP CORRELATIONS
January 2006
Published by
ENERGY INSTITUTE, LONDON
The Energy Institute is a professional membership body incorporated by Royal Charter 2003
Registered charity number 1097899
The Energy Institute gratefully acknowledges the financial contributions towards the scientific and
technical programme from the following companies:
BG Group
BHP Billiton Limited
BP Exploration Operating Co Ltd
BP Oil UK Ltd
Chevron
ConocoPhillips Ltd
ENI
ExxonMobil International Ltd
Kuwait Petroleum International Ltd
CONTENTS
Page
Foreword..vii
Acknowledgements.......ix
Overview....xi
SECTION 1 - Phase 1 Report - Ignition probability review and model development .................................... 1
1.1
Introduction ............................................................................................................................................... 1
1.2
Literature review ....................................................................................................................................... 2
1.2.1
Review of historical ignition probability data............................................................................. 2
1.2.2
Ignition models and related issues .............................................................................................. 6
1.3
HCR OIR12 ignition event analysis........................................................................................................ 13
1.3.1
OIR12 2001 ignition data.......................................................................................................... 13
1.3.2
OIR12 2000 leak and ignitions data.......................................................................................... 17
1.3.3
OIR12 analysis - Summary ....................................................................................................... 24
1.4
Ignition modelling and related topics...................................................................................................... 24
1.4.1
Gas dispersion within plant areas.............................................................................................. 24
1.4.2
Gas dispersion outside plant areas ............................................................................................ 26
1.4.3
Ignition probability methods..................................................................................................... 27
1.4.4
Self-ignition .............................................................................................................................. 30
1.4.5
Ignition modelling and dispersion summary............................................................................. 31
1.5
Review conclusions and implications for ignition modelling ................................................................. 32
1.5.1
Model approach ........................................................................................................................ 32
1.5.2
Ignition data and simple probability correlations...................................................................... 32
1.5.3
Ignition timing and fire-explosion probabilities for use in QRA .............................................. 33
1.5.4
Release source modelling.......................................................................................................... 34
1.5.5
Control factors .......................................................................................................................... 35
1.5.6
Ignition timing and explosion probabilities Model implications ........................................... 35
1.5.7
Categorisation of fluid types ..................................................................................................... 36
1.5.8
Model development Initial concept........................................................................................ 36
1.6
Model description.................................................................................................................................... 38
1.6.1
Model structure ....................................................................................................................................... 38
1.6.2
Input data .................................................................................................................................. 41
1.6.3
Model output............................................................................................................................. 50
1.6.4
Model process detailed description........................................................................................... 51
1.6.5
Workbook model overview....................................................................................................... 63
1.7
Model benchmarking............................................................................................................................... 64
1.8
Model conclusions and recommendations .............................................................................................. 66
1.9
Issues for future work.............................................................................................................................. 66
Annex
Annex 1.A References........................................................................................................................................ 67
Figures
Figure 1.1 Simple correlation of fraction of liquid not raining out vs. discharge pressure based on RELEASE
model estimates ................................................................................................................................... 10
Figure 1.2 Simple correlation of fraction of liquid not raining out vs. discharge pressure based on cyclohexane
test data................................................................................................................................................ 11
Figure 1.3 Cumulative event based ignition probabilities - based on average release rate................................... 21
Figure 1.4 Cumulative event based ignition probabilities - based on maximum release rate ............................... 21
Figure 1.5 Ratio between maximum (initial) release rate and averaged release rate for 294' cut of OIR12 data 22
Figure 1.6 Cumulative event distribution for 294 data cut................................................................................. 22
Figure 1.7 Gas build-up JIP workbook - Output for selected typical module configurations .............................. 25
Figure 1.8 A comparison of different indicators for gas dispersion...................................................................... 26
Figure 1.9 A comparison of ignition probabilities assigned by various models for some selected scenarios....... 30
Figure 1.10 Ignition modelling Areas.................................................................................................................... 37
Figure 1.11 Ignition modelling Areas.................................................................................................................... 39
Figure 1.12 Alternative ignition modelling Areas approach ................................................................................. 39
Figure 1.13 Ignition model structure ..................................................................................................................... 40
Figure 1.14 Ignition modelling Areas.................................................................................................................... 44
Figure 1.15 Ignition modelling Areas Cloud areas............................................................................................. 45
Figure 1.16 Alternative ignition modelling Areas approach ................................................................................. 45
Figure 1.17 Gas build-up JIP workbook correlation ............................................................................................. 57
Figure 1.18 Liquid release aerosol formation correlation ..................................................................................... 59
Figure 1.19 Release maximum vs. average release rate Decay factor................................................................ 61
Figure 1.20 Model comparison Gas releases...................................................................................................... 65
Figure 1.21 Model comparison Liquid releases ................................................................................................. 65
Tables
Table 1.1 Analysis of offshore ignition sources ...................................................................................................... 2
Table 1.2 Generic ignition probabilities .................................................................................................................. 2
Table 1.3 Generic explosion probabilities (given ignition) ..................................................................................... 2
Table 1.4 Ignition sources for industrial plant......................................................................................................... 3
Table 1.5 Generic ignition probabilities (CMPT).................................................................................................... 3
Table 1.6 Generic ignition probabilities (Cox et al) ................................................................................................ 3
Table 1.7 Ignition delay probabilities...................................................................................................................... 3
Table 1.8 Judged ignition probabilities for a range of offshore scenarios............................................................... 4
Table 1.9 Proposed generic ignition probabilities for typical offshore hydrocarbon releases................................. 4
Table 1.10 Suggested method for assigning ignition timings.................................................................................. 5
Table 1.11 Ignition timings - Based on an analysis of LPG releases ...................................................................... 5
Table 1.12 Probability of explosion given ignition - Based on offshore data ......................................................... 5
Table 1.13 Onshore gas pipeline leak ignition probabilities ................................................................................... 6
Table 1.14 Onshore gasoline pipeline leak ignition probabilities ........................................................................... 6
Table 1.15 Ignition source densities (per hectare)................................................................................................... 8
Table 1.16 Ignition parameters for area based model ........................................................................................... 9
Table 1.17 Estimated rain-out fractions for flammable liquid releases (RELEASE model) ................................... 9
Table 1.18 Analysis of all ignitions by cause and source of ignition .................................................................... 13
Table 1.19 Analysis of OIR12 MAH related ignition events ................................................................................ 14
Table 1.20 Analysis of OIR12 database large releases.......................................................................................... 17
Table 1.21 Analysis of OIR12 database large releases - more bands.................................................................... 18
Table 1.22 Analysis of OIR12 database large releases - 499 events by maximum and average release rate
basis ................................................................................................................................................. 19
Table 1.23 Analysis of OIR12 database large releases - comparison with cumulative basis ................................ 20
Table 1.24 Analysis of large liquid releases.......................................................................................................... 23
Table 1.25 Analysis of OIR12 database - Ignition timing..................................................................................... 23
Table 1.26 Analysis of OIR12 database - Ignited events ...................................................................................... 24
Table 1.27 CHAOS based model from DNV TDIM JIP - Appendix G - Gives average gas concentration in
module vs. time ................................................................................................................................... 25
Table 1.28 DNV TDIM - Recommended ignition source densities ...................................................................... 27
Table 1.29 Application of DNV TDIM ignition sources to a typical module ....................................................... 27
Table 1.30 Selected results for a typical offshore module based on the WS Atkins on-site ignition
methodology ........................................................................................................................................ 28
Table 1.31 Comparison of ignition probabilities assigned by various models to selected scenarios .................... 29
Table 1.32 Self-ignition probabilities.................................................................................................................... 31
Table 1.33 Generic ignition probabilities (Cox et al and Revised)..................................................................... 32
Table 1.34 Ignition timings overview.................................................................................................................... 33
Table 1.35 Relative incidences of ignited events .................................................................................................. 33
Table 1.36 Pool spill sizes ..................................................................................................................................... 35
Table 1.37 IP fluid categories................................................................................................................................ 36
Table 1.38 Relationship between IP fluid categories and petroleum classes ........................................................ 36
Table 1.39 Ignition model input data..................................................................................................................... 41
Table 1.40 Ignition area types set up in model...................................................................................................... 43
Table 1.41 Release types - Data entry ................................................................................................................... 49
ii
iii
Annex 2.B Ignition modelling results figures and look-up correlation figures for scenario specific
configurations ................................................................................................................................. 107
Figures
Figure 2.1 Ignition modelling Areas................................................................................................................... 70
Figure 2.2 Ignition model structure .................................................................................................................... 70
Figure 2.3 Model results for onshore pipeline Oil......................................................................................... 108
Figure 2.4 Model results for onshore pipeline LPG ...................................................................................... 108
Figure 2.5 Model results for onshore pipeline Gas ....................................................................................... 109
Figure 2.6 Model results for small onshore plant Open plant area................................................................ 109
Figure 2.7 Model results for small onshore plant Confined/ congested plant area........................................ 110
Figure 2.8 Model results for large onshore plant Open plant area ................................................................ 110
Figure 2.9 Model results for large onshore plant Confined/ congested plant area ........................................ 111
Figure 2.10 Model results for onshore tank farm LPG Passive dispersion ................................................ 111
Figure 2.11 Model results for onshore tank farm Oil Passive dispersion................................................... 112
Figure 2.12 Model results for onshore tank farm LPG Release treated as momentum jet ......................... 112
Figure 2.13 Model results for onshore tank farm Oil Release treated as momentum jet............................ 113
Figure 2.14 Model results for offshore process module Oil releases ............................................................ 113
Figure 2.15 Model results for offshore wellhead Oil releases....................................................................... 114
Figure 2.16 Model results for offshore NUI..................................................................................................... 114
Figure 2.17 Model results for offshore FPSO process module ........................................................................ 115
Figure 2.18 Model results for offshore FPSO process module behind transverse wall................................. 115
Figure 2.19 Model results for offshore process module Gaseous releases.................................................... 116
Figure 2.20 Model results for offshore wellhead and compression modules Gaseous releases .................... 116
Figure 2.21 Model results for offshore riser Release near to edge of installation ......................................... 117
Figure 2.22 Model results for offshore riser Release away from edge of installation................................... 117
Figure 2.23 Model results for offshore release engulfing platform.................................................................. 118
Figure 2.24 Cox, Lees and Ang ignition probabilities for comparison ............................................................ 118
Figure 2.25 Look-up vs. model results for offshore NUI................................................................................. 119
Figure 2.26 Look-up vs. model results for offshore process module Gas ..................................................... 119
Figure 2.27 Look-up vs. model results for offshore process module Gas Congested or mechanically
ventilated module.......................................................................................................................... 120
Figure 2.28 Look-up vs. model results for offshore risers ............................................................................... 120
Figure 2.29 Look-up vs. model results for offshore FPSO no wall............................................................... 121
Figure 2.30 Look-up vs. model results for offshore FPSO with transverse wall .......................................... 121
Figure 2.31 Look-up vs. model results for offshore release engulfing platform .............................................. 122
Figure 2.32 Look-up vs. model results for offshore process module Liquid ................................................ 122
Figure 2.33 Look-up vs. model results for onshore pipelines Gas/ LPG ...................................................... 123
Figure 2.34 Look-up vs. model results for onshore pipelines Liquid............................................................ 123
Figure 2.35 Look-up vs. model results for onshore small plant Gas/ LPG ................................................... 124
Figure 2.36 Look-up vs. model results for onshore small plant Liquid......................................................... 124
Figure 2.37 Look-up vs. model results for onshore large plant Gas/ LPG .................................................... 125
Figure 2.38 Look-up vs. model results for onshore large plant Liquid ......................................................... 125
Figure 2.39 Look-up vs. model results for onshore storage Gas/ LPG ......................................................... 126
Figure 2.40 Look-up vs. model results for onshore storage Liquid............................................................... 126
Figure 2.41 Look-up correlations (best fit) for onshore liquid releases ........................................................... 127
Figure 2.42 Look-up correlations (best fit) for onshore gas/LPG releases....................................................... 127
Figure 2.43 Look-up correlations (best fit) for offshore liquid releases........................................................... 128
Figure 2.44 Look-up correlations (best fit) for offshore gas releases............................................................... 128
Figure 2.45 Look-up correlations (best fit) for offshore engulfing release and riser release ........................... 129
Tables
Table 2.1 Ignition model input data.................................................................................................................... 71
Table 2.2 Ignition area types set up in model..................................................................................................... 73
Table 2.3 Release materials and properties ........................................................................................................ 75
Table 2.4 Release mass flow rates used.............................................................................................................. 75
Table 2.5 Scenarios pressures used .................................................................................................................... 78
Table 2.6 Offshore module f4 geometry and orientation factor assessment....................................................... 79
Table 2.7 Storage ignition parameters ................................................................................................................ 80
Table 2.8 Ignition probability scenario configurations....................................................................................... 86
iv
vi
FOREWORD
The probability of ignition of flammable releases from onshore and offshore installations is a key factor in
determining the risks such installations present to people, assets and the environment. Yet, data and practical
models available to enable assessors to assign suitable ignition probability distributions in quantified risk
analysis (QRA) have been lacking.
IP Ignition probability review, model development and look-up correlations provides the findings of a
United Kingdom Offshore Operators Association (UKOOA)/ Health and Safety Executive (HSE)/ Energy
Institute (EI) co-sponsored project undertaken by ESR Technology (formerly the Engineering Safety and Risk
Business of AEA Technology) to improve the modelling of ignition probabilities in onshore and offshore
installation QRAs. Specifically, the project reviewed current data and models, developed an ignition
probability model for assigning ignition probabilities in QRA, and developed a simple basis and guidance for
assigning ignition probabilities to generic scenarios.
Phase 1 of the project reviewed knowledge and research in the field of ignition probability estimation
in support of QRA. This was evaluated, together with the usefulness of the UK offshore industry hydrocarbon
release data, as a basis for developing an improved ignition probability model. The review concluded that the
commonly applied approach of adopting generic correlations based on the mass release rate for the probability
of ignition was overly simplistic, and may lead to unrealistic and very conservative estimates in some situations.
Further, some of the correlations in use do not reflect recent ignition probability data and knowledge.
An ignition probability model was developed that combined recent work on the dispersion of
flammable vapours in process areas with the identification of ignition characteristics in offshore and onshore
plant areas and other relevant onsite and offsite land use types. It has been benchmarked against historical
ignition probability data.
The model made available as a functional workbook provides a means to estimate the overall
ignition probability and an approximate time/ location distribution for specific release scenarios. It can also
provide an insight into the main ignition factors and allows sensitivity analysis and 'what-if' analysis, which
may help designers to change the plant layout or process conditions to reduce the ignition potential.
Recognising that the number of factors taken into account by the model means it may be too onerous to
apply for some risk assessments, Phase 2 of the project involved development of a series of simple, mass release
rate based, ignition probability look-up correlations for a selected range of representative onshore and offshore
plant and storage release scenarios. These correlations made available as a functional workbook provide
users with a simple and convenient introductory means to incorporate more appropriate ignition probability data
in many typical QRAs. Their scope includes liquid and vapour release correlations; in addition, two-phase
releases can be assessed by combining the liquid and vapour results. The look-up correlations are supported by
guidance on selecting the appropriate correlation to use in a given situation, and details of the underlying basis
and assumptions used in deriving them. Note that the full model should be run for complex scenarios or where
additional detail is required.
Some guidance is provided on ignition timing and estimating the likelihood of an explosion given
ignition.
Note that the model and look-up correlations are intended to assess the ignition potential of momentum
driven releases of hydrocarbons; however, they are not suited to the ignition probability assessment of
refrigerated liquefied gases, vaporising liquid pools, sub-sonic gas releases, or non-momentum driven releases,
such as those following catastrophic storage vessel failure.
IP Ignition probability review, model development and look-up correlations should be drawn to the
attention of safety practitioners, risk analysts and designers who need to determine or evaluate ignition
probabilities in QRA, such as where required for safety cases or safety reports.
The project sponsors believe that the model and look-up correlations provide a more appropriate means
for assigning ignition probabilities in QRA as they take better account of the release type and plant
characteristics, and results better align with historical data including the UK offshore industry hydrocarbon
release data. However, their use requires expert judgement and results should be used and interpreted with a
degree of caution, in particular where they are less conservative than former approaches. In addition, those
preparing or revising safety cases or safety reports should include details of pertinent assumptions and input
data that support the derived ignition probabilities.
The information contained in this IP Research Report is provided as guidance only and while every
reasonable care has been taken to ensure the accuracy of its contents, the Energy Institute, the project cosponsors, the developer and the technical representatives listed in the acknowledgements, cannot accept any
vii
responsibility for any action taken, or not taken, on the basis of this information. The Energy Institute shall not
be liable to any person for any loss or damage which may arise from the use of any of the information contained
in any of its publications.
The information contained in this IP Research Report may be reviewed from time to time. It would be
of considerable assistance in any future revision if users would send comments (e.g. on the performance of the
model and the look-up correlations) or suggestions for improvement to:
The Technical Department
Energy Institute
61 New Cavendish Street
London W1G 7AR
e: technical@energyinst.org.uk
viii
ACKNOWLEDGEMENTS
The project was undertaken by David Mansfield ESR Technology1 (formerly the Engineering Safety and Risk
Business of AEA Technology) and co-ordinated by a joint industry steering group drawn from the project
sponsors, petroleum operating companies and senior risk assessment consultants working in the field of onshore
and offshore QRA:
David Aberdeen
Brian Bain
Gerry Burke
Stephen Connolly
Howard Crowther
Tim Davenport
Alastair Gunn
Richard Jones
Robert Kyle
Colin Morgan
Mark Scanlon
Chris Venn
BP (Chairperson)
DNV
WS Atkins
Health and Safety Executive
Consultant
Shell
ExxonMobil
Acona
United Kingdom Offshore Operators Association
ESR Technology
Energy Institute
WS Atkins
The Institute wishes to record its appreciation of the work carried out by them and others who participated
during the technical review, for example by testing the draft deliverables.
Affiliations refer to the time of participation.
Co-sponsorship of this project by UKOOA and HSE is gratefully acknowledged.
ESR Technology, Whittle House, 410 The Quadrant, Birchwood Park, Warrington, WA3 6FW, United Kingdom.
Telephone +44 (0)1925 843400. Facsimile +44 (0)1925 843500
ix
OVERVIEW
This IP Research Report comprises the following project deliverables:
Phase 1 report: Ignition probability review and model development (provided as part of this Adobe
Acrobat document) (issued to project sponsors as AEAT/NOIL/27780001/001(R) Issue 1);
UKOOA ignition model workbooks (Version D1) in single scenario and multirun formats (provided as
Microsoft Excel workbooks) (issued to project sponsors as Ukooa_Ignition_model_D1 Version D1 and
Ukooa_Ignition_model_D1_multirun Version D1, respectively);
Phase 2 report: Look-up correlations (provided as part of this Adobe Acrobat document) (issued to
project sponsors as AEAT/NOIL/28129001/002(R) Issue 1);
Look-up correlations workbook (Version D1) (provided as a Microsoft Excel workbook) (issued to
project sponsors as Ukooa_pign_lookup_function_D1 Version D1).
Given that the project was carried out in two phases, some aspects of the Phase 1 report have been superseded
by the Phase 2 report. In addition, the workbooks and Phase 1 report refer to earlier versions of the workbooks.
Furthermore, the process of preparing the project deliverables for publication necessitated some editorial
changes to the versions issued to the project sponsors. Notwithstanding this, none of these issues should affect
the technical integrity of this IP Research Report.
Phase 1 report: Ignition probability review and model development
The Phase 1 report reviews knowledge and research in the field of ignition probability estimation in support of
QRA. This was evaluated, together with the usefulness of the UK offshore industry hydrocarbon release data, as
a basis for developing an improved ignition probability model. The review concluded that the commonly
applied approach of adopting generic correlations based on the mass release rate for the probability of ignition
was overly simplistic, and may lead to unrealistic and very conservative estimates in some situations. Further,
some of the correlations in use do not reflect more up-to-date historical ignition probability data and knowledge.
The report describes the basis for the ignition probability model, which combined recent work on the
dispersion of flammable vapours in process areas with the identification of ignition characteristics in offshore
and onshore plant areas and other relevant onsite and offsite land use types. The model has been structured to
consider the ignition of hydrocarbons within the immediate area where the leak occurs, and any additional
probability of ignition were the flammable vapour cloud or liquid to spread to adjacent areas or beyond.
The model can provide an insight into the main ignition factors and allows sensitivity analysis and
'what-if' analysis, which may help designers to change the plant layout or process conditions to reduce the
ignition potential.
The report also describes benchmarking of the model against historical ignition probability data.
UKOOA Ignition model workbooks
The ignition probability model, which is referred to as the UKOOA Ignition model is provided in single
scenario and multirun formats as Microsoft Excel workbooks (Version D1); the main difference being that the
latter also includes a macro to run the model for a variety of user defined scenarios and release rates.
The model input data consists of basic information on the release, the four Areas being considered
and their ignition characteristics, and a number of options for modelling the release and any subsequent
dispersion beyond the immediate area of release.
The model provides an estimate of the overall probability of ignition for the scenario, together with
contributions to the probability from vapour and liquid components of the release for each of the four Areas
represented in the model. It also provides an approximate time distribution of the ignition probability as it
spreads across them.
The Microsoft Excel workbooks also include user notes and change records.
Phase 2 report: Look-up correlations
Recognising that the number of factors taken into account by the ignition probability model means it may be too
onerous to apply for some risk assessments, the Phase 2 report describes use of the model to prepare a series of
xi
mass release rate based, ignition probability look-up correlations for a selected range of representative onshore
and offshore plant and storage release scenarios. These correlations provide users with a simple and convenient
introductory means to incorporate more appropriate ignition probability data in many typical QRAs. Their
scope includes liquid and vapour release correlations; in addition, two-phase releases can be assessed by
combining the liquid and vapour results. The look-up correlations are supported by guidance on selecting the
appropriate correlation to use in a given situation, and details of the underlying basis and assumptions used in
deriving them.
However, users are cautioned that there will be situations where the nature or complexity of the
scenario or the detail of the assessment means that the look-up correlations will not be sufficient. In these cases
the full model should be used to estimate the relevant ignition probabilities.
The Phase 2 report also offers guidance on ignition timing and estimating the likelihood of an
explosion given ignition. In addition, it describes several improvements which were made to the model,
primarily to provide additional error traps and to better represent plant storage area ignition characteristics.
Look-up correlations workbook
A Microsoft Excel workbook (Version D1) enables the ignition probability to be determined from the look-up
correlations using a simple cell function formula requiring mass flow rate and scenario type inputs, linked to a
visual basic module in the workbook.
The workbook also contains other worksheets with information and useful supporting features related
to the look-up correlations, such as a help guide to selecting the most suitable release type for a given situation,
and the basis and characteristics of the correlations. Its functionality enables users to view all look-up
correlation charts, those of similar types (e.g. onshore plant) or user-defined combinations.
xii
INTRODUCTION
This phase of the project aims to provide a technical review of ignition data and methods to develop improved
models and guidance for the estimation of ignition probabilities for use in quantified risk analysis (QRA). It is
recognised that current methods can be very simplistic and do not take account of recent leak and ignition data,
which are showing a reduced trend in ignition probabilities. Current models may also not take into account
other factors known to be important in ignitions such as vapour dispersion and the concentration of ignition
sources. The study reviews the current data and models and develops improved models based on the current
level of understanding. It also suggests areas for future investigation and development, which could be
addressed in a Phase 2 project.
Ignition probabilities are key inputs in QRA that can have a major influence on the estimated risks. To
date, the ignition probabilities adopted for offshore and many onshore QRAs have been drawn from relatively
few sources and data points, many of which may not be appropriate to the modern design and operation of oil
and gas installations.
Attempts have been made in recent years to develop simple models or relationships to assign ignition
probabilities in QRAs. A common approach to assigning ignition probability is to relate this to the media and
release flow rate. Such an approach is proposed in the E&P Forum Hydrocarbon leak and ignition database[6]
and in the Cox, Lees and Ang formulae[2], which were developed in 1990.
These formulae typically assign a minimum ignition probability of 1 % for small releases rising to a
maximum of 30% for 100 kg/s releases. The upper limits have generally been derived from offshore blowout
ignition data. This assumes that the maximum ignition probability for any offshore release would be no greater
than the average ignition probability from a range of blowouts on drilling units; an assumption that does not
stand up to detailed interrogation.
Models and simple fractions have also been suggested to estimate what proportion of ignitions are early
or delayed and which may result in explosions (an example is given in Cox, Lees and Ang) [2]. These
probabilities determine the hazardous outcome from any ignited event (e.g. jet fire, flash fire or explosion
followed by a jet fire), and so can be critical in determining the risk characteristics.
Given the greater attention to major accident hazard (MAH) safety, and the need for ignition control
highlighted by the current safety regimes in the UK North Sea and the European oil and gas industry in general,
it could be expected that current ignition probabilities might be less than those from past experience. This may
be being reflected in some of the latest ignition data.
The objective of this phase of the project is to provide a current and informed view on ignition
probabilities for use in QRA, and develop a simple model to assist practitioners in assigning ignition
probabilities for use in QRA, together with some guidance on the model and its application. The study
addresses the timing of ignition and methods for estimating the likelihood of an explosion, given ignition. The
work includes detailed reviews of the latest OIR12 leak and ignitions data and recent blowout ignition data.
The model development needed to consider both offshore and onshore situations where practicable,
although the main focus is on the offshore situation. There are key differences between these environments
including the confined layout of most offshore installations as opposed to the relatively open nature of many
onshore facilities. It is considered that this would have little effect on the nature and timing of the ignition of
small leaks. These will be limited to the module/ plant area concerned and any ignition sources within this.
The difference will be with large leaks that could spread to adjacent areas and reach external sources of ignition
- the leak needed to spread from one offshore module to the next would be smaller than ones capable of drifting
between onshore plant areas. The study attempts to estimate what size of leaks for onshore vs. offshore could
spread to adjacent modules/ reach external ignition sources and factor this into the ignition probability model.
It was envisaged that the overall ignition probability model, where practicable, should include the rule
sets or correlations for ignition timing and the likelihood of an explosion. Guidance on the model and rules
sets/ correlations would need to be provided.
Given the uncertainty in estimating ignition probabilities, ignition timing and explosion probabilities, it
is essential that the models, rule sets or correlations developed by the project are set out and justified, and that
guidance is provided to assist the user to select the most appropriate factors when applying these in a specific
QRA. It was envisaged that the guidance would also include, where appropriate, typical values and upper and
lower bounds for factors or probabilities against a range of common scenarios. This is to assist users in the
selection of these factors and in undertaking sensitivity or uncertainty analysis as part of the QRA.
1.2
LITERATURE REVIEW
1.2.1
A review of some of the more important data sources has been undertaken and key points noted.
1.2.1.1 Forsth
Survey identifies causes of offshore fires and explosions, Forsth (Det Norske Veritas), Technology Oil and Gas
Journal, 25 July 1983, pp 154-158 [1].
This reviews some 140 fires and explosions in Gulf of Mexico and 133 in Norwegian North Sea reported on
OCS Accident Investigation and Analysis Reports. Welding and hot surfaces are shown to be the predominant
ignition sources. See Table 1.1.
Table 1.1 Analysis of offshore ignition sources
Ignition source
Welding/cutting/grinding
Engines and exhausts
Sparks
Electrical
Hot surfaces (excluding engines and
exhausts)
Self-ignition
Cigarette/lighter/match
Other
Unknown/not reported
Total
1.5
1.5
5
12
100
1.2.1.2
Cox et al
Classification of Hazardous Locations, Cox, Lees and Ang, Institute of Chemical Engineers, May 1991, ISBN 0
85295 258 9 [2]
This presents a review of onshore and offshore ignition data and suggests a simple ignition model based on
mass release rate to assist in risk based calculations to support area classification. Simple log-log plots are
presented based on the anchor points given in Tables 1.2-1.3.
Table 1.2 Generic ignition probabilities
Release rate category
Minor
Major
Massive
Gas leak
0.01
0.07
0.3
Oil leak
0.01
0.03
0.08
Data are also presented on the sources of ignition for onshore industrial plant. See Table 1.4. Those data have
also been analysed to provide data for open and closed process plant, but this specifically excludes spontaneous
ignition and hot work since the compilers viewed this as usually both the source of the leak and the ignition.
No.
8
2
10
4
8
3
6
7
38
%
9.3
2.3
11.6
4.7
9.3
3.5
7.0
8.1
44.2
No.
2
27
20
11
29
10
17
2
21
%
1.4
19.4
14.4
7.9
20.9
7.2
12.2
1.4
15.1
86
100.0
139
100.0
1.2.1.3 CMPT
A guide to quantitative risk assessment for offshore installations, J Spouge, CMPT, 1999, Publication 99/100,
ISBN 1 870553 365 [3]
Reviews historical ignition probabilities based on the ratio of leaks to ignitions from a variety of offshore
sources. Generic ignition probabilities for use in QRA are suggested based on the material concerned (oil or
gas) and the mass flow rate. See Table 1.5. The large leak data is based on SINTEF blowout incident data.
The Cox, Lees and Ang 1990 data [2] are also presented as an alternative. See Table 1.6. A set of judgmental
data are also presented in Table 1.7 to estimate the timing of any ignition.
Table 1.5 Generic ignition probabilities (CMPT)
Release rate category
Tiny
Small
Medium
Large
Massive
Gas leak
0.005
0.04
0.10
0.30
0.5
Oil leak
0.03
0.04
0.06
0.08
0.10
Gas leak
0.01
0.07
0.3
Oil leak
0.01
0.03
0.08
The timing data in Table 1.7 are said to be based largely on judgement and to align reasonably well with
blowout data. However, they do not align with data from the OIR12 database on process leaks, or the onshore
3
data presented in Table 1.11, or the blowout data from Per Holand[5] which indicate 40% immediate ignition and
a further 15 % within 5 min. to 1 hr.
1.2.1.4 HSE OSD research
A review of ignition probabilities for use in offshore installation quantified risk assessments, AEA Technology,
HSE OSD Research Project 3209, May 1995 [4]
Review of both onshore and offshore ignition data and models. Proposed a range of ignition probabilities based
on the limited data available and judgement for a range of typical offshore situations taking account of the
release size, material and the degree of confinement (see Table 1.9).
Typical scenarios in order of judged decreasing likelihood of ignition (major leaks) are given in
Table 1.8.
Table 1.8 Judged ignition probabilities for a range of offshore scenarios
No.
A
B
C
D
E
F
G
H
I
J
Scenario
Gas leak in module - poor ventilation
Gas leak in module - good ventilation
Oil leak in module - poor ventilation
Gas release on drilling rig
Oil leak in module - good ventilation
Oil release on drilling rig
Gas leak on open deck
Subsea/air gap gas release
Oil spill on open deck
Subsea/air gap oil release
Probability of ignition
0.15
0.05
0.08
0.05
0.05
0.05
0.03
0.03
0.025
0.025
Table 1.9 Proposed generic ignition probabilities for typical offshore hydrocarbon releases
Fluid
Gas/ condensate
Situation
Release rate
Confined
Massive
Major
Minor
Massive
Major
Minor
Massive
Major
Minor
Massive
Major
Minor
Massive
Major
Minor
Massive
Major
Minor
Semi confined
Open
Oil
Confined
Semi confined
Open
Nominal ignition
probability
0.5
0.15
0.05
0.3
0.05
0.01
0.2
0.03
0.0025
0.15
0.08
0.05
0.08
0.05
0.03
0.05
0.025
0.01
Ignition timing data and historical ratios of fires to explosions were also reviewed (see Tables 1.10-1.12).
These concluded that data were sparse but indicated that between 0.3 and 0.7 of leaks ignited early (within
30 s). The chance of an explosion given ignition varied from 0.15-0.2 for process gas leaks, and between 0.30.4 for blowouts. A proposed judgmental interpretation of data is presented to assign ignition timing for a
variety of plant situations taking account of whether hot work is carried out in the unit and the degree of
congestion. Data are also presented on an analysis of onshore releases (mainly of LPG) by M Considine
(HSE/SRD PDD50WP2).
Plant type
Plant
Transport
Pipelines
Overall
1
0.22
0.53
0.24
0.33
> 1000
1.0
1.0
1.0
1.0
The report also quotes data from the World Offshore Accident Databank (WOAD), Statistical Report 1994 on
the incidence of fires and explosions.
Table 1.12 Probability of explosion given ignition - Based on offshore data
Event
Fire
Explosion
Probability of explosion
World-wide data
343
95
0.22
The report also quotes blowout data indicating a 36 % chance of an explosion given ignition of a gas blowout.
No explosions were recorded for oil blowouts or oil/ gas mixture blowouts. Two other data sources indicated
explosion probabilities of approximately 0.15 given ignition.
1.2.1.5 Offshore blowouts Per Holand
Offshore blowouts, causes and trends, Per Holand, Norwegian Institute of Technology, Trondheim, Norway
1996 [5]
Per Holand reviews world-wide blowout data and provides estimates of the frequency of blowouts for a variety
of conditions. This indicates ignition probabilities of approximately 15 % (depending on the type of blowout)
with approximately 37.5 % occurring immediately, 15 % delayed but within 5 min.-1 hr., and 47.5 % over an
hour or more - in fact these were over 6 hours after the leak commenced! However, it is noted that the
likelihood of ignition of blowouts seems to be falling, presumably as better ways of working and controls are
introduced. As a result, Per Holand recommends an average ignition probability of 0.1 for blowouts in QRAs.
1.2.1.6 E&P Forum
Hydrocarbon leak and ignition database, Report No 11.4/180, May 1992, E&P Forum [6]
This reviews limited offshore historical ignition data and uses them to anchor a simple ignition model based on
the volume of the gas/ vapour cloud.
The probability of ignition is given by:
P ign = 1 - exp (-0.005m1.5)
This is based on the 0.005 data of the ignition probability for a small leak. However this equation gives an
upper ignition probability of 1.0 for large leaks. As a result the equation was factored using the ignition
probability for a blowout (P max) to set this as its upper limit:
P ign = 1 - exp (-0.005m1.5) x P max
1.2.1.7 EGIG
4th EGIG Report 1970-1998 Gas pipeline incidents, Doc. No. EGIG 99.R.0074 December 1999 [7]
This gives historical derived ignition probabilities for onshore gas transmission pipelines, with hole sizes
classified by the hole diameter. See Table 1.13.
Table 1.13 Onshore gas pipeline leak ignition probabilities
Hole type
Pinhole-crack
Hole
Rupture 16 inch pipe diam.
Rupture > 16 inch pipe diam.
Average for all leaks
Probability of ignition
0.034
0.017
0.105
0.286
0.038
Rupture
0.031
0.0155
0.0155
Hole
0.031
0.0155
0.0155
Leak
0.0062
0.0031
0.0031
Rupture
0.062
0.031
0.031
Hole
0.062
0.031
0.031
Leak
0.0124
0.0062
0.0062
inside a confined or congested plant area. It should be noted that the method is based on naturally ventilated
modules; it may not be suitable for modelling confined areas with forced ventilation.
1.2.2.2 WS Atkins - HSE research Phase 1
Ignition probability of flammable gases, WS Atkins, HSE Contract Research Report 146/1997, HSE Books,
1997, ISBN 0 7176 1431 X [10]
The report reviews ignition data and models for use in onshore risk assessment. A model is proposed to assess
the ignition probability for drifting vapour clouds over off-site areas. The model takes account of the likely
ignition sources in the area and the time the vapour cloud takes to pass these within its flammable limits. It also
has a feature to enable gas ingress to buildings to be taken into account so that the influence of internal ignition
sources can be more accurately represented.
Data are also presented on possible ignition sources and their characterisation in terms of ignition
potential, density per unit area of land, and rate of activity (continuous or intermittent).
The first part of the Phase 1 WS Atkins report contains a review of ignition modelling and available
data to 1997. The references quoted are as given in the Atkins report[10].
This notes that simple ignition probability models used extensively in QRAs during the 1990s tended
to be based on either the very small amount of incident data available or expert judgements, as reviewed by both
Cox et al (1990) and CCPS (1995).
Expert judgement over the years has produced estimates based on a single figure or a few bands. For
example, Browning (1969) suggests that, for massive LPG leaks into areas with no obvious source of ignition,
and explosion proof equipment, the probability of ignition is only 0.1. Kletz (1977) argued that the probability
of ignition increases with size of leak and is certainly greater than 0.1 for large leaks (10 ton or more) and may
be as high as 0.5. The Canvey Island Report (HSE, 1981) used onsite ignition probabilities ranging from 0.1
for areas with no ignition sources to 0.9 for many ignition sources.
The approach adopted by Blything & Reeves (1988) was that 70 % of large LPG gaseous releases
would ignite. A large release was defined as one reaching in excess of about 60 m to LFL. For other release
sizes, ignition probability was then taken to be proportional to distance travelled. Ignition probability was
reduced by a factor of ten for cases where the releases did not reach the nearest identifiable ignition source.
Rew et al (1998) suggest that this model, which was compared against LPGITA data, underpredicts by a factor
of ten for small releases, possibly as a result of ignoring temporary ignition sources, e.g. hot work.
There is wide variation in the values of ignition probabilities given by the various authors quoted
above. Many of the values are case specific and may not be directly applicable to more general studies.
However, the models that have been commonly used are often based on little more than judgement as
the ignition data are so sparse. The models built up using simple correlations for ignition probability have been
based on historical data rather than site information. Examples based on cloud area are Clay et al, 1988 and
Simmons, 1974, whilst the commonly used Cox et al, 1990 uses release rate.
Simmons (1974) was based on data from a survey of 59 incidents of ignition of clouds of LNG or LPG
resulting from accidental spills due to transportation. For these, the size of the cloud when ignition occurred
was estimated and fitted the probability of ignition as a function of cloud area to an error function.
An onshore offsite mathematical ignition probability model is also presented that uses a combination of
release dispersion and ignition source characteristics. Main dependencies used in the model are cloud size,
concentration and ignition source distribution and type. Ignition source time dependence and multiple ignition
source modelling are carried out.
1.2.2.3 WS Atkins - HSE Research Phase 2
A model for the ignition probability of flammable gases (Phase 2), WS Atkins, HSE Contract Research Report
203/1998, HSE Books, 1998, ISBN 0 7176 1657 6 [11]
This continues work reported in the Phase 1 WS Atkins report - developing and testing a model to assess the
ignition probability of drifting vapour clouds from onshore MAH facilities.
The Phase 2 WS Atkins report contains the industrial, rural and urban ignition data from which
ignition source parameters have been developed. An offsite ignition model is then developed. This is then
compared with existing models (e.g. HSE model, Clay et al, 1988).
The model uses a grid system to address the various land use and ignition source characteristics within
the path of the vapour cloud. The model determines the time the cloud takes to pass over the various ignition
sources, and hence the chance of ignition within the time window.
The ignition sources are characterised by:
The ignition potential, p, the basic probability of ignition given the source is active and in contact with
flammable gas.
The ignition source density, , which gives the number of sources per unit area (per hectare)
For intermittent sources:
The time the source is active, ta (min.)
The time period between each activation, ti (min.)
From these values the frequency (per min.) at which the source becomes active, , and the probability of the
source being active, a, can be calculated:
= 1 / (ta + ti)
The overall ignition probability at time t is then calculated based on the following summation for the probability
of non-ignition at time t, Q(t):
ln Q(t) = I = 1 to I j = 1 to J Ai j [ (1 - aj pj ) e -j pj di - 1]
Where:
A is the area of the ith region being considered in contact with the gas cloud (hectares)
j = 1 to j represents the jth ignition source
d is the duration the source has been in contact with the flammable gas cloud (min.)
Only offsite ignition is addressed. The model is not intended for use within plant areas.
Outputs are provided that could be used to develop simple correlations between the cloud area (at
ground level) and the probability of ignition for different land use conditions (day, night, industrial, urban,
rural). Generic ignition source densities are also presented (Table D.2) for these different conditions based on
the model currently in use by the HSE (see Table 1.15).
Table 1.15 Ignition source densities (per hectare)
Day
Night
The report also contains aggregated ignition characteristics for some typical industrial activities and plant,
which could be used in a simplified area ignition model.
1.2.2.4 WS Atkins - HSE Research Phase 3
Development of a method for the determination of on-site ignition probabilities, WS Atkins, HSE Research
Report 226, HSE Books, ISBN 0 7176 2843 4, 2004 [12]
A further extension of the WS Atkins work looks at the development of an onshore onsite ignition model based
on gas dispersion, specific ignition source locations and the effect of controls on potential ignition sources
(typically hazard area classification and permit to work).
The model assumes that Zone 1 and 2 hazardous areas do not contribute to the ignition potential.
A 12 hectare LPG tank storage facility is considered as an example. The site is split into its different
land use regions (e.g. car park, boiler house, etc.). The gas cloud shape and range is derived using the code
Hegabox. Potential ignition sources are represented using a time dependent model and given a location.
The conclusions are that poor ignition control (hazardous area classification and use of permit to work)
would lead to an ignition probability approaching 1 in the event of a catastrophic release. Typical levels of
ignition control result in probabilities building up to 0.9 but not reaching about 1 until the cloud area more than
exceeds the size of the site. For a good level of ignition control, the ignition probability is still considered to
reach 0.9.
The report includes estimates of the ignition source parameters for typical industrial activities,
including process plant with light, medium and heavy equipment levels, utility areas, etc. (see Table 1.16, based
on Table E.2 of the WS Atkins report). These are used in an area based model to assess the ignition of a
drifting vapour cloud over the site.
Typical
equipment
Large motors,
pumps, etc.
Smaller
motors and
pumps, etc.
Low voltage
switchgear
and contacts
only
Ignition
potential
Activation
time fraction
Frequency
of activation
(min.)
High
Medium
Low
0.5
0.0277
50
25
12.5
0.3
0.0347
50
25
12.5
0.15
0.0556
50
25
12.5
The assessment method assumes that Zone 1-2 hazardous areas present a negligible ignition potential.
Ventilation rates are used to assess the build-up of gas in an enclosed building to assess the potential
for ignition inside plant buildings.
The sample results presented in the report all indicate that the probability of ignition increases quickly
with time, typically rising to 60 % of its ultimate value within 10 s and 90 % within 60 s.
1.2.2.5 IP Classification and rain-out
Calculations in support of IP15: The area classification code for petroleum installations, Institute of Petroleum,
November 2001, ISBN 0 85293 339 8 [13]
This presents work on two-phase releases, in particular the generation of mist/ aerosol from liquid releases at
pressure. The analysis suggests that at higher pressures, all the release can be considered to remain airborne.
Table 1.17 is presented as being typical of a Class C liquid (flammable liquid) and the data are calculated using
the RELEASE model[14]. The IP report acknowledges that these estimates are not necessarily correct, but does
point out that expert review of the RELEASE model indicates that it is conservative, overestimating the amount
of rain-out. The report concludes that rain-out should be ignored for pressurised releases when calculating
hazard radii. For example test data from the cyclohexane trials at 350 K used to develop the RELEASE model
indicate 50 % rain-out at pressures as low as 2.1 bara, 33 % at 2.8 bara and 9.5 % at 3.9 bara respectively (see
Figure 1.2) (RELEASE A model with data to predict aerosol rainout in accidental releases, D Johnson and J
Woodward, CCPS, AIChE, 1999, ISBN 0-8169-0745-5).
Table 1.17 Estimated rain-out fractions for flammable liquid releases (RELEASE model)
Pressure (bara)
5
10
50
100
The data in Table 1.17 can be represented by an approximate s-curve shaped correlation (see Figure 1.1):
Fraction not raining out, F = If p<7.5, then F = 10^((3/8.3)*p-3.1),
else if (LN(p/1.854)/3.11>1, then F = 1, else F = LN(p/1.854)/3.11))
where p is the fluid pressure in barg.
Figure 1.1 Simple correlation of fraction of liquid not raining out vs. discharge pressure based on
RELEASE model estimates
1
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
0
10
15
20
25
30
35
40
45
50
Pressure, barg
However these general model predictions should be treated with caution. In practice, the degree of rain-out
depends on both the superheat of the material being released and the potential for mechanical break-up as
indicated by the Weber number. It may be more appropriate for the modelling of source term issues such as
rain-out to be done outside the ignition model. The situation would also be complicated by the shape of the
release site, which is also likely to have a significant effect - contrast a flange leak with an accidentally open
valve. The presence of equipment in the area could also cause further evaporation and droplet break-up in the
liquid jet/ spray. A conservative approach may be needed to account for these uncertainties.
Using the actual test data suggests higher fractions in the air, and can be represented by the following
correlation:
Fraction not raining out, F = if p<0.5, then F = p*0.286/0.5,
else F = 1-EXP(-0.15*(1+p)^2)
where p is the fluid pressure in barg.
This may provide a more conservative correlation in terms of aerosol/ mist generation for use in the ignition
probability model.
10
Figure 1.2 Simple correlation of fraction of liquid not raining out vs. discharge pressure based on
cyclohexane test data
1
0.9
0.8
Fraction in Air
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
0
10
Pressure, barg
11
12
1.3
An analysis of the 2000 OIR12 data (from the data diskette) has been undertaken together with additional
analysis of the published 2001 Ignition data extracted from Offshore hydrocarbon release statistics, 2001, HID
Statistics Report HSR 2001 002, HSE, January 2002 [16].
1.3.1
An analysis of the ignition events in the OIR data-set up to 2001 is presented in Table 1.18.
Table 1.18 Analysis of all ignitions by cause and source of ignition
Ignition source
Turbine engine hot surface exhaust
Hot work
Flare-vent
Turbine internal
Glycol reboiler
Static - manual ops
Non process related
Drilling
Unknown
Electrical fault
Total
No.
62
27
13
11
5
4
4
2
2
1
131
%
47.3
20.6
9.9
8.4
3.8
3.1
3.1
1.5
1.5
0.8
100.0
An analysis of the ignition events was undertaken to identify those that could be considered of relevance to the
MAH assessment. Each event was marked as not relevant (N), possibly relevant (P) and relevant (Y) on the
basis of the nature of the leak and ignition. For example incidents inside turbine enclosure and vent ignitions by
snow/ lightning were discounted, as were events involving non-process fluids. The findings of this screening
process are presented in the Table 1.19. The incidents are grouped by process media type, and are in the order
of increasing release rate-based on a simple pressure/ hole size correlation. A discussion of the analysis of these
MAH related events follows.
The MAH findings show that hot work is by far the most significant cause of ignition (22 out of 30
events). Four events are attributed to manual operation or static - mainly due to container handling, poor
earthing strap connections or use of non anti-static equipment. Two events involved a series of fires and
explosions in a flare system and were included as possibles because they were different from the more
common vent ignitions. Only one event was related to (faulty) electrical equipment - the type of ignition
scenario associated with classic MAH release ignition. Finally, there was one ignition of unknown source,
which may have been static.
The analysis also shows that almost all the ignitions occurred during discrete operating modes construction, inspection, maintenance, sampling, etc. It should be noted that the operating mode construction
does not relate to major construction work but is used to denote rebuilding and replacement work and smallscale construction work associated with an operating installation, and as such may be relevant to the ignition of
production releases. Only two events were associated with normal production; one of the flare incidents and
the electrical fault.
The majority of the possible MAH related ignitions involved small quantities of material two thirds of
the gas ignitions were classed as minor releases and all the oil and condensate releases were classed as
minor.
The analysis suggests that few, if any, of the ignition events could be taken to be representative of
typical MAH release scenarios, i.e. significant process leaks during operation igniting as the released material
dispersed to reach ignition sources.
13
Drains, Closed
Minor
Minor
78
87
14
Export, Oil
Significant
Significant
Significant
80
88
26
Vent, LP
Minor
Minor
54
53
Utilities, Oil, Heat
Transfer Oil
Manifold, Oil
116 Minor
Vent, HP
Metering, Oil
Drains, Closed
103 Minor
Minor
86
Drilling, Development,
Gas Well, <100
Export, Oil
Minor
115 Minor
79
Vent, LP
63
Minor
105 Minor
Import, Oil
Gas Compression
System
130 Minor
Gas
Ref. Severity
0.1
0.1
0.09
41.1
0.9
0.9
0.8
0.8
0.5
0.1
0.1
0.07
0.03
10
0.01
0.14
Quantity Pressure
released (barg)
(kg)
10
0.25
0.5
0.5
0.5
0.1
0.25
Release
duration
(min.)
3.9
152.4
12.7
203.2
457.2
25.4
9.1
9.1
N/A
25
N/A
4.1
Hole
diam.
(mm)
Sampling
Operating mode
Delay
time (s)
Hot work
Hot work
Hot work
Hot work
Hot work
Hot work
Unknown
Hot work
Hot work
Hot work
Hot work
Hot work
Static
Manual Ops
Hot work
Hot work
15
117 Minor
Significant
51
Utilities, Gas, Fuel Gas
Flare, HP
Processing, Gas,
LPG/Condensate
Significant
Flowlines, Gas
62
Minor
22
Processing, Gas,
LPG/Condensate
Export, Gas
Significant
89
System
129 Significant
Minor
23
Gas (continued)
Ref. Severity
0.01
40.5
14.4
1.8
0.6
121.9
0.03
75.86
165
0.06
2.62
0.79
0.69
300
49
103.45
0.25
Release
duration
(min.)
12.41
Quantity Pressure
released (barg)
(kg)
N/A
N/A
12.7
Hole
diam.
(mm)
Normal Production
Normal Production
Inspection
Operating mode
60
9999
Delay
time (s)
Hot work
Flare Vent
Hot work
Hot work
Hot work
Elec Fault
Hot work
16
Minor
Minor
Minor
50
77
Minor
49
21
Condensate
Import, Gas
Flare, HP
Processing, Gas,
LPG/Condensate
Export, Oil
127 Minor
Processing, Gas,
Dehydration
Minor
19
Export, Oil
System
101 Minor
Minor
Oil
Ref. Severity
0.7
5.99
10
27
0.2
3.63
47.9
Quantity Pressure
released (kg) (barg)
0.5
Release
duration
(min.)
25.4
N/A
12.7
2.2
N/A
12.7
508
Sampling
Cleaning
Delay
time (s)
Hot work
Static
Flare Vent
Hot work
Hot work
Static
Static
Hot work
1.3.2
An analysis of the data provided in the data diskettes up to and including the year 2000 data has been
undertaken to try to identify any common factors or ignition probabilities that could be used to anchor an
ignition model.
In particular, previous analyses of the data have shown that no major classified leaks have ignited to
date. An analysis was undertaken of the larger leaks in the database to try to establish if any common factors
e.g. ventilation, shutdown, were accounting for the lack of ignitions.
An analysis of all significant and major process leaks with maximum release rates above 0.5 kg/s
and averaged leak rates above 0.2 kg/s was carried out. This identified 294 events of which six ignited.
Analysis of the six ignition events showed that four of these related to flare ignitions i.e. of oil or condensate
carry over, and the other two were fuel gas ignitions. None related to the ignition of process fluids from the
main process.
The individual event records show that the unignited events ranged in size and duration, with a range
of ventilation (natural, forced, not known, various windspeeds, some as low as 2 m/s) and shutdown (auto,
manual, none) characteristics. No specific trends could be identified.
In fact the ignition data to date point to events that occur due to flare problems, utility turbine/ exhaust
problems and hot work igniting small leaks and spills. It is likely that many of these were only reportable
because they ignited. The overall conclusion seems to be that none of the releases in the database, which one
would associate with an offshore process QRA, have ignited.
An indication of the probability of ignition for use in a QRA can be gained by counting the number of
unignited leaks so far recorded in the database that are of a type considered in a typical QRA. This is discussed
in 1.3.2.1.
1.3.2.1
Analysis by maximum and average release rate
Based on the 294 events selected in 1.3.2, releases were categorised by leak size and type.
Table 1.20 Analysis of OIR12 database large releases
Release type
No. events
Pign (Note 2)
No. events
Pign (Note 2)
Gas
>50 kg/s
3
0.125
21
0.023
1-50 kg/s
54
0.009
130
0.004
<1 kg/s
94
0.005
0
0.500
Oil
>50 kg/s
1
0.250
14
0.033
1-50 kg/s
26
0.019
56
0.009
<1 kg/s
51
0.010
0
0.500
Condensate
>50 kg/s
0
0.500
6
0.071
1-50 kg/s
11
0.042
17
0.028
<1 kg/s
15
0.031
0
0.500
2-Phase (Note 1)
>50 kg/s
0
0.500
4
0.100
1-50 kg/s
13
0.036
24
0.020
<1 kg/s
26
0.019
0
0.500
Note 1: 2-Phase refers to multiphase well fluids prior to first stage separation
Note 2: Ignition probability estimate assuming 50 % chance that next leak ignites
0.019
0.002
0.008
14
74
15
0.033
0.007
0.031
6
20
8
0.071
0.024
0.056
6
46
11
0.071
0.011
0.042
Using the averaged release rate (quantity released divided by the duration) gives lower flow rates than the
maximums calculated using standard orifice flow equations for the initial release conditions. As a result there
are fewer events in the high flow categories in the first data column of Table 1.20. The maximum (initial) flow
rate may be more appropriate for assigning QRA ignition probabilities and for classifying the data for use in
QRA. The average flow rate was primarily included in the filtering of the events to exclude very small leaks,
i.e. where a very low flow rate persisted for some time, or where the initial conditions would give a high flow
rate, but the limited inventory meant that the leak quickly exhausted itself.
17
An indication of the upper bound of an overall ignition probability for each leak category is estimated
assuming that there is be a 0.5 chance that the next leak ignites.
For the large gas leaks, this indicates an ignition probability of 0.02 or less. The leaks of oil,
condensate and multiphase fluids would be expected to have ignition potentials at least that of a general
flammable liquid. Taking these events together gives a large liquid leak ignition probability of 0.02 or less.
Leaks in the mid-range (1-50 kg/s) suggest an ignition probability of approximately 0.01 for gas and
0.02 for liquids.
Small leaks exhibit similar values, 0.01 for gas and 0.04 for liquids.
These values are in line with overall ignition rates estimated from more global analyses of the OIR12
data, indicating ignition probabilities of a few percent.
Table 1.21 Analysis of OIR12 database large releases - more bands
Basis average release rate
Gas
>50
>20
>10
>5
>2
>1
>0
Oil
>50
>20
>10
>5
>2
>1
>0
Condensate
>50
>20
>10
>5
>2
>1
>0
2-phase
>50
>20
>10
>5
>2
>1
>0
No.
leaks
No.
ignitions
Pign (using
data or next
leak 50 %
chance if no
ignitions)
Pign (on
No. leaks
No.
basis next
ignitions
leak 50 %
chance of
ignitions,
ignores any
actual
ignitions)
Pign (using
data or next
leak 50 %
chance if no
ignitions)
50
20
10
5
2
1
3
3
6
4
17
24
94
0
0
0
0
0
0
2
0.125
0.125
0.071
0.100
0.028
0.020
0.021
0.125
0.125
0.071
0.100
0.028
0.020
0.005
22
16
23
30
47
11
2
0
0
0
1
0
1
0
0.022
0.029
0.021
0.033
0.010
0.091
0.167
0.022
0.029
0.021
0.016
0.010
0.042
0.167
50
20
10
5
2
1
1
3
3
1
8
11
51
0
0
0
0
0
0
3
0.250
0.125
0.125
0.250
0.056
0.042
0.059
0.250
0.125
0.125
0.250
0.056
0.042
0.010
14
4
8
8
26
14
4
0
0
0
0
0
0
3
0.033
0.100
0.056
0.056
0.019
0.033
0.750
0.033
0.100
0.056
0.056
0.019
0.033
0.100
50
20
10
5
2
1
0
1
1
3
3
2
16
0
0
0
0
0
0
1
0.500
0.250
0.250
0.125
0.125
0.167
0.063
0.500
0.250
0.250
0.125
0.125
0.167
0.029
6
2
3
3
9
2
1
0
0
0
0
0
0
1
0.071
0.167
0.125
0.125
0.050
0.167
1.000
0.071
0.167
0.125
0.125
0.050
0.167
0.250
50
20
10
5
2
1
0
1
0
1
7
4
26
0
0
0
0
0
0
0
0.500
0.250
0.500
0.250
0.063
0.100
0.019
0.500
0.250
0.500
0.250
0.063
0.100
0.019
4
3
7
4
6
10
5
0
0
0
0
0
0
0
0.100
0.125
0.063
0.100
0.071
0.045
0.083
0.100
0.125
0.063
0.100
0.071
0.045
0.083
18
A further analysis based on the 499 event cut was also undertaken to assess any difference between criteria
based on the maximum (initial) release rate and the average release rate for each leak event. See Table 1.22.
1.3.2.2
Analysis by seven release rate bands
Analysis of the 294 events using a more narrow set of cuts was also undertaken to see if this provided a better
delineation of events in the higher release rate categories. See Table 1.21.
Table 1.22 Analysis of OIR12 database large releases - 499 events by maximum and average release
rate basis
Release rate criteria
(kg/s)
Gas
>50
1
0
Oil
>50
1
0
Condensate
>50
1
0
2-phase
>50
1
0
No. leaks
No. leaks
50
<1
2
60
237
0.167
0.008
0.002
25
205
69
0.019
0.002
0.007
50
<1
1
26
76
0.250
0.019
0.006
14
69
20
0.033
0.007
0.024
50
<1
0
10
24
0.500
0.045
0.020
6
19
9
0.071
0.025
0.050
50
<1
0
15
48
0.500
0.031
0.010
6
44
13
0.071
0.011
0.036
Note: There are some slight differences between this analysis and that in Table 1.20 in the Draft A Phase 1
report issued to the Joint Industry Steering Group these are due to the way zero maximum release rate events
were dealt with in the original manual analysis. These are treated as zero above; previously they were treated as
per the average release rate. The differences are not significant.
1.3.2.3
Cumulative analysis
Given that some of the release rate bands selected have few events in them, an alternative analysis was
performed on the 294 event cut using a cumulative approach. See Table 1.23. The ignition probabilities are
based on a 50 % chance of the next leak igniting with the relevant data cut taken from all events above the
specified release rate criteria. The analysis ignores any actual ignitions of releases in each category since the
few ignitions that are recorded were considered to not be of a type relevant to MAH risk assessment.
The cumulative ignition probabilities are presented graphically in Figure 1.3 and Figure 1.4.
The results support the broad conclusion that average ignition probabilities are approximately 2 % for
gas and perhaps a little higher (2-5 %) for some liquids. The sparse data for the higher release rates (>20 kg/s)
mean that care should be taken interpreting the figures. The indications are that average ignition probabilities
for these types of releases are unlikely to exceed 0.1 (10 %). It could also be argued that the data for
condensate, oil and 2-Phase (i.e. multiphase media prior to separation - typically low GOR oil) releases are all
essentially liquid type releases, and so these data can be combined to provide an overall indication of the
average liquid ignition probability. This would bring the average ignition probability for small to medium sized
releases of liquids down to approximately 2 %, in line with the gas figure.
19
Table 1.23 Analysis of OIR12 database large releases - comparison with cumulative basis
on basis next leak
has 50 % chance of
ignition, ignores any
actual ignitions
Release rate criteria
(kg/s)
Pign
Pign based on
cumulative distribution
of release events i.e. >
stated release rate
Pign based on
cumulative distribution
of release events i.e. >
stated release rate
0.125
0.125
0.071
0.100
0.028
0.020
0.005
0.125
0.071
0.038
0.029
0.015
0.009
0.003
0.022
0.029
0.021
0.016
0.010
0.042
0.167
0.022
0.013
0.008
0.005
0.004
0.003
0.003
0.250
0.125
0.125
0.250
0.056
0.042
0.010
0.250
0.100
0.063
0.056
0.029
0.018
0.006
0.033
0.100
0.056
0.056
0.019
0.033
0.100
0.033
0.026
0.019
0.014
0.008
0.007
0.006
0.500
0.250
0.250
0.125
0.125
0.167
0.029
0.500
0.250
0.167
0.083
0.056
0.045
0.019
0.071
0.167
0.125
0.125
0.050
0.167
0.250
0.071
0.056
0.042
0.033
0.021
0.019
0.019
0.500
0.250
0.500
0.250
0.063
0.100
0.019
0.500
0.250
0.250
0.167
0.050
0.036
0.013
0.100
0.125
0.063
0.100
0.071
0.045
0.083
0.100
0.063
0.033
0.026
0.020
0.014
0.013
0.250
0.083
0.100
0.083
0.026
0.028
0.005
0.250
0.071
0.045
0.031
0.015
0.010
0.003
0.020
0.050
0.026
0.031
0.012
0.019
0.045
0.020
0.015
0.010
0.007
0.005
0.004
0.003
Gas
>50
>20
>10
>5
>2
>1
>0
Oil
>50
>20
>10
>5
>2
>1
>0
Condensate
>50
>20
>10
>5
>2
>1
>0
2-phase
>50
>20
>10
>5
>2
>1
>0
All liquids combined
>50
>20
>10
>5
>2
>1
>0
20
Figure 1.3 Cumulative event based ignition probabilities - based on average release rate
0.500
0.450
0.400
GAS
OIL
0.350
CONDENSATE
2-PHASE
0.300
0.250
0.200
0.150
0.100
0.050
0.000
0
10
20
30
40
50
60
Figure 1.4 Cumulative event based ignition probabilities - based on maximum release rate
0.200
0.180
GAS
OIL
0.160
CONDESATE
2-PHASE
0.140
0.120
0.100
0.080
0.060
0.040
0.020
0.000
0
10
20
30
40
50
60
1.3.2.4
Analysis of leak characteristics
The data were also investigated to assess how characteristic they were of typical process leaks. An analysis
based on the maximum release rate to average release rate ratio was undertaken. The maximum release rate
(kg/s) was calculated using standard orifice equations and the event actual pressure and hole size. The average
release rate (kg/s) was calculated by dividing the inventory released by the release duration (converted to
seconds). The data were also analysed by presenting the cumulative distribution of events vs. release rate, to
21
indicate the relative quantities of events of a given release rate or above. The results are presented in Figure 1.5
and Figure 1.6.
Figure 1.5 Ratio between maximum (initial) release rate and averaged release rate for 294 cut of OIR12
data
1.2
0.8
0.6
0.4
0.2
0
0.1
10
100
1000
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
1
10
100
22
1000
The results show that the majority of releases have the typical characteristics of a release of locked in material at
pressure, i.e. show maximum to average release rate ratios of between 2 and 20. The events showing very low
maximum to average ratios are probably small weeping leaks where the release persisted for some time, but at a
greatly reduced flow rate compared to the initial release rate or predicted theoretical flow rate. The step at the
top of the plot is due to the events where no hole size was specified, in these cases the analysis set the maximum
rate equal to the average rate giving a ratio of 1.0. There are also a few releases where the maximum to average
ratio is less than 1.0. These could be due to conservative assumptions in the data reported or the release rate
calculations, e.g. for 2-phase releases undertaken to support the analysis. However, the number of events of this
type is insignificant in terms of the overall analysis.
The release rate distribution (based on the 294 event cut) shows that most releases had average
release rates below 10 kg/s and initial/ maximum release rates of below 200 kg/s. Nevertheless the analysis
shows that many of the leaks are of a size that would be relevant to a MAH risk analysis, especially in the small
to mid range release sizes. The relatively few events exhibiting massive release rates (above 100 kg/s) mean
that any conclusions from the data may not be applicable to the very large releases considered at the upper end
of a risk assessment.
1.3.2.5
Analysis of large liquid releases (rain-out)
An analysis of the large liquid leaks from the 294 data cut was undertaken to investigate if it were possible to
establish any evidence of mist/ aerosol formation when these releases were at pressure. In general, it is
anticipated that high pressure releases of liquid through certain types of orifice can generate substantial
quantities of aerosol/ mist, perhaps leading to almost full entrainment (i.e. with minimal rain-out of droplets) in
the air where the pressure is above 10 barg. The analysis looked at oil, condensate and 2-phase leaks. The
results are presented in Table 1.24.
Table 1.24 Analysis of large liquid releases
Pressure criteria
>0 bar
>5 bar
>10 bar
>20 bar
% detected by gas
detection
39
43
45
45
Most of the 2-phase releases had low GORs (approx. 90 % less than 0.5, 60 % less than 0.1) and were
associated with oil systems e.g. oil separation, oil flow-line. This suggests most were primarily liquid media
and justifies their inclusion in this liquid analysis. The incidence of detection of the leaks by gas detection
was similar for all operating pressure bands. The data unfortunately cannot be further interrogated to determine
if this is simply a function of the gas detection philosophy, or whether the higher pressure releases are not
generating significantly greater quantities of mist/ aerosol/ vapour. The analysis is therefore inconclusive.
1.3.2.6
Ignition timing and end events analysis
The timing of ignition and the nature of the ignited events were also investigated using the OIR12 2000 data.
Table 1.25 Analysis of OIR12 database - Ignition timing
Timing
Delayed
Immediate
Total
No. of events
3
48
51
% of ignited events
5.88
94.12
Comment
2 within a minute
The results in Table 1.25 suggest that most leaks ignited immediately. This needs to be considered against the
context that many of these leaks were minor events involving hot work or flare liquid break-through in which
case early ignition would be likely.
23
Flash fire
29
56.86 %
Explosion
Jet fire
All process leaks
6
10
11.76 %
19.61 %
Oil leaks
5
71.43 %
3
37.50 %
21
58.33 %
Pool fire
6
11.76 %
2
28.57 %
Condensate leaks
1
12.50 %
Gas leaks
5
10
13.89 %
27.78 %
4
50.00 %
0
0.00 %
Analysing the same events by the resulting ignited event type indicates that the incidence of explosions is
twice that of delayed ignitions. See Table 1.26. This means that the traditional QRA approach to assume
delayed ignitions equate to explosions may be misleading. Further, if we postulate that flash fires could present
an explosion potential if confined, then the incidence of events with explosion potential increases to
approximately 60 % of all ignited events. As mentioned above, the data need to be considered in the context of
the type of leaks that ignited; few of these would be representative of the accidental leaks considered in a MAH
QRA.
1.3.3
Analyses of both the OIR12 2001 ignition data and the OIR12 2000 raw data indicate that most, if not all
ignitions to date would be of little relevance to MAH risk analysis. Most involved the ignition of small leaks or
spills during manual activities such as hot work or sampling, or internal turbine events, flare carry-over events
and fuel leaks on to hot exhausts in utility areas. These would not be relevant to the ignition of accidental
process leaks during production activities. An investigation into events with significant leak rates showed that
none of the ignitions associated with these events were relevant to QRA.
The best estimates of overall ignition probabilities, therefore, need to be derived from the data on
events that have not ignited. This enables a crude estimate of the upper limit of the ignition probabilities. The
analysis indicates that gas leaks in the 1-50 kg/s and >50 kg/s ranges have ignition probabilities of 0.02 or less.
Liquid releases are similar. For small leaks, the gas ignition rate falls to 0.02 or less, whereas there exists the
possibility that liquid releases may exhibit a higher ignition rate of 0.04 or less. This apparent increase in
ignition probabilities for smaller liquid releases may be a function of the reporting criteria rather than a true
reflection of the ignition potential.
The OIR12 data have also been reviewed to try to establish if any factors can account for the lack of
ignition, particularly of the larger releases. The analysis showed that large leak events cover a range of different
situations, ventilation configurations and rates, shutdown responses, etc. No common factors could be
identified, other than the fact that MAH type inadvertent process leaks so far do not appear to have been ignited
- indicating that current design and operational controls are reasonably effective.
1.4
1.4.1
A comparison has been undertaken of various methods to assess the dispersion of gas within plant areas and, in
particular, to establish when the area becomes saturated with gas.
24
Figure 1.7 Gas build-up JIP workbook - Output for selected typical module configurations
100.00%
10.00%
1.00%
Enclosed
Fully Open
Open Ended
0.10%
Free jet
0.01%
0.00%
0.1
10
100
1000
Figure 1.7 shows that typically, the volume of a flammable cloud reaches a saturation value of approximately
60 % once the mass flow rate reaches a few kg/s, even in a relatively open module configuration. This means
that the method predicts that all but the smallest releases will quickly fill the module with gas within the
flammable limits only occupying half the volume. The volume occupied by a free turbulent jet to LFL (based
on a simple correlation) is also included for comparison. (Module size taken as 20 x 10 x 6 m.)
The volume of a simple momentum jet also highlights the possible shortcomings of the Gas Build-up
JIP Workbook method with very small pressurised low flow rate releases. Here the volume of the gas jet could
be more dependent on the momentum jet rather than the ventilation in the area (i.e. a fixed volume that would
not vary with the degree of confinement).
The CHAOS based methodology used in the DNV JIP TDIM method[15] calculates an average gas
concentration within a given plant area. See Table 1.27. Unfortunately this gives no indication of the volume
of the plant area which is exposed to gas within flammable limits at any given point in time. However, it can
indicate the point at which the ventilation either results in a global dilution below LFL or saturation above UFL.
Table 1.27 CHAOS based model from DNV TDIM JIP - Appendix G - Gives average gas concentration
in module vs. time
Mass flow rate (kg/s)
0.1
0.2
0.5
1
2
5
10
20
50
100
60
60
60
60
60
60
60
60
60
60
25
Gas concentration
(vol/vol)
0.006
0.012
0.029
0.057
0.112
0.260
0.464
0.748
1.106
1.226
1
1200
6.6
600
These results show that once the mass flow reaches approximately 1 kg/s, the module average concentration is
above LFL and once above approximately 3 kg/s the average concentration is above UFL.
Figure 1.8 A comparison of different indicators for gas dispersion
100
Jet length = module 'length'
Enclosed Module at 60%
Open Module at 60%
10
0.1
0.01
CASE 1 = 10 by 3 by
3 module
Figure 1.8 shows three different approaches that could be used to indicate when an area becomes saturated with
gas. Results are presented for four module configurations. The module average concentration at LFL is given
by the CHAOS based method in the DNV TDIM workbook[15]. The Module 60 % values are given by the Gas
Build-up JIP Workbook method[9] and represent the flow rate at which the module just reaches 60 % of volume
with flammable gas (i.e. saturation). The jet length = module length method simply equates the free
momentum jet length to the averaged module dimension (the radius of the equivalent volume sphere). The
results show that for a typical module (open ended), the transition to a saturated gas regime could be predicted
by the Gas Build-up JIP Workbook method or the simple jet length correlation. These also equate to the point
at which the CHAOS method predicts an average LFL concentration in the module.
It is also interesting to note the relatively low flow rates that are required to saturate the modules;
typically of the order of 1 kg/s. This suggests that even small inventories of gas or flashing liquid could sustain
these conditions for several minutes; sufficient for the gas to reach any ignition sources that might be present in
the area, and perhaps to warm up against hot surfaces or allow time for intermittent sources such as faulty
electrical equipment to generate sparks.
1.4.2
The ignition method developed by Atkins for the HSE[12] approaches the estimation of external ignition
probabilities by the use of an area based method, looking at the land area that is exposed to the flammable gas
cloud. However, no correlations are given in the method to estimate the size of the cloud. The external ignition
method (XDIS) developed by DNV in its TDIM model[15] uses a similar approach with correlations based on
26
CFD modelling of gas dispersion around (enclosed) platforms/ installation modules. It allows the concentration
at some specific point from the release to be estimated. Different correlations are provided depending on the
configuration of the release/ platform/ ignition source location of interest. Since the method is location specific,
it does not lend itself to a more general approach based on the cloud size and ignition source density.
Simple correlations are also published in the literature to assess the hazard range and area of drifting
gas clouds. For example Rapid Assessment of the Consequences of LPG Releases, M Considine and GC
Grint, Gastech 1984, gives correlations for continuous and quasi-instantaneous releases of both pressurised and
refrigerated flammable gases. Suitable correlations of this type could be selected from existing literature or
derived from a range of representative modelling runs to estimate the hazard range and area/ volume within the
cloud.
It is also possible to adapt these correlations to take account of the available inventory, truncating the
dispersion from a continuous release plume based on the mass of material within the cloud. This would be one
way of accounting for the effects of isolation and blowdown on dispersion and hence, the ignition probability.
1.4.3
The DNV TDIM and WS Atkins methods have been applied to selected module/ plant configurations to
compare these methods. See Tables 1.28-1.29.
Table 1.28 DNV TDIM - Recommended ignition source densities
Source
Unit
Continuous
Gas
Discrete (per s)
Gas
Oil
Oil
3.30 x 10-6
5.70 x 10-5
3.40 x 10-8
5.80 x 10-7
-5
-3
-7
3.40 x 10-6
Pump
per item
6.40 x 10
4.40 x 10
Compressor
per item
1.50 x 10-3
1.50 x 10-2
2.70 x 10-6
2.70 x 10-5
Generator
per item
3.50 x 10
-3
-2
-6
6.20 x 10-5
Other
per m2
1.40 x 10-5
8.40 x 10-9
1.50 x 10-7
Personnel
per m2
5.50 x 10-6
2.60 x 10-4
2.80 x 10-8
(From Appendix K, Table 8.31 Recommended ignition source intensities)
5.80 x 10-7
3.50 x 10
4.10 x 10-4
1.20 x 10
6.20 x 10
Description
Gas
Basic module with no
5.40 x 10-3
specific high ignition risk
equipment
Module with 3 compressors 1.05 x 10-2
and 2 pumps
Module with 3 generators
1.70 x 10-2
Ignition probability
averaged per unit
volume of module,
(m3)
Gas
Oil
Oil
Gas
Oil
1.61 x 10-1
2.70 x 10-5
8.06 x 10-4
4.50 x 10-6
1.34 x 10-4
2.20 x 10-1
5.27 x 10-5
1.10 x 10-3
8.78 x 10-6
1.83 x 10-4
2.77 x 10-1
8.51 x 10-5
1.39 x 10-3
1.42 x 10-5
2.31 x 10-4
As the whole module ignition probabilities show, the probability of ignition estimated is very low. The data
presented above take no account of the mass release rate or the area/ volume of the module exposed to
flammable gas, so these values could be interpreted as very conservative. In the actual DNV TDIM model,
these ignition intensities are combined with information on the volume of the module exposed to flammable gas
concentrations and the time of the exposure to estimate the ignition probability.
27
Table 1.30 Selected results for a typical offshore module based on the WS Atkins on-site ignition
methodology
Scenario
Basis 20 x 10 x 6 m module
Area , Ignition
(m2 of
source
cloud
density
to
(per ha)
LFL)
a, prob.
source
active
p,
ignition
potential
, ignition
frequency
(per min.)
Exposure
duration
(s)
P ign
120
50
5.7 x 10-3
0.5
1.1 x 10-4
10
1.7 x 10-3
120
50
0.1
5.6 x 10-2
5.8 x 10-2
120
50
0.1
5.6 x 10-2
5.8 x 10-2
50
0.1
5.6 x 10-2
10
1.8 x 10-3
0.04
50
0.1
5.6 x 10-2
30
2.1 x 10-4
50
0.5
5.6 x 10-2
2.6 x 10-1
3.6
50
0.5
5.6 x 10-2
10
9.0 x 10-3
50
0.5
5.6 x 10-2
30
1.1 x 10-3
Note that the ignition sources for the light to heavy process areas in Table 1.30 are based on unzoned areas,
i.e. with no area classification. The Atkins methodology assumes zoned areas have no ignition potential.
These results can be compared with the DNV TDIM results given in Table 1.31.
28
Table 1.31 Comparison of ignition probabilities assigned by various models to selected scenarios
Scenario
Basis 20 x 10 x 6 m module
P ign by
P ign by
P ign by
P ign by
Area (m2 DNV TDIM
DNV TDIM WS Atkins DNV TDIM Cox, Lees
of cloud
Ignition
and Ang
to LFL) Intensity (per Method
Method
Method
m2)
(Area basis) (Area basis) (Volume
(Flow
basis)
basis)
120
2.85 x 10-5
3.42 x 10-3
1.7 x 10-3
5.71 x 10-3
120
2.70 x 10
-5
-3
-2
-3
3.6
3.24 x 10
5.8 x 10
3.24 x 10
2.44 x 10-2
2.70 x 10-5
9.72 x 10-5
1.8 x 10-3
1.05 x 10-5
1.56 x 10-2
0.04
2.70 x 10-5
1.08 x 10-5
2.1 x 10-4
3.33 x 10-7
3.56 x 10-3
120
8.51 x 10-5
1.02 x 10-2
2.6 x 10-1
1.02 x 10-2
2.44 x 10-2
3.6
8.51 x 10-5
3.06 x 10-4
9.0 x 10-3
3.32 x 10-5
1.56 x 10-2
0.04
8.51 x 10-5
3.40 x 10-5
1.1 x 10-5
1.05 x 10-6
3.56 x 10-3
Note: 60 % of full module cases - Cox, Lees and Ang ignition probability based on a mass release rate of
2 kg/s - the flow rate indicated by the JIP Gas Build-up Workbook as at the transition point at which an open
ended module would become saturated with gas. Other volumes based on simple momentum jet
representation.
The results in Table 1.31 are presented graphically in Figure 1.9.
29
Figure 1.9 A comparison of ignition probabilities assigned by various models for some selected scenarios
1
WS Atkins Method -area basis
DNV TDIM Method - area basis
DNV TDIM Method - volume basis
0.1
Probability of Ignition
0.001
0.0001
0.00001
0.000001
0.0000001
Hot work - 50
hrs/year
Light equip in Light equip in Light equip in Heavy equip in Heavy equip in Heavy equip in
process area - process area - process area - process area - process area - process area 60% of full
1 kg/s medium 0.1 kg/s small
60% of full
1 kg/s medium 0.1 kg/s small
leak, so long
module, short
leak, so
leak, so long
module, very
leak, so
short contact medium contact contact time contact time for medium contact contact time
gas to reach
time
time for gas to
time
LFL
reach LFL
Comparing the value shows that the DNV method gives ignition probabilities approximately 20 - 30 times
lower than the WS Atkins method, except for the hot work ignition probability which is of the same order. It is
to be expected that the DNV method would give a lower probability for equipment in modules, since this is
intended to reflect equipment in zoned areas, whereas the WS Atkins method takes zoned areas to be
negligible in terms of ignition potential and the data are based on unclassified equipment. Both methods
generally give lower values than the Cox, Lees and Ang mass release rate based method.
Using a volume based approach reduces the ignition probability for small and medium sized leaks
significantly, since the volume ratio is much smaller than the floor area ratio. Although the volume based
approach may be more realistic, it should be noted that the ignition potentials for both the WS Atkins and
DNV TDIM methods have been allocated on an area basis. The DNV TDIM ignition intensity values were
converted to a volume basis for the volume ignition estimates above, but this is not the intended method of
application.
Hot work ignition probabilities, based on the typical fraction of time hot work is undertaken in a
process plant, give ignited event frequencies much lower than those reported in databases such as the OIR12
database. This is probably due to a number of factors including:
The hot work activity may also have been the cause of the leak.
The incident may have been very minor and was only reported because it ignited.
The leak or build-up of hydrocarbon could have been an ongoing problem that was only detected when
ignited by the hot work.
Care should therefore be applied when interpreting historical ignition data on hot work to the ignition of MAH
type releases.
1.4.4
Self-ignition
Recent ignition modelling work based on a modified version of the DNV TDIM model by Lilleaker Consulting
(the ASAP risk package) includes an estimate for self-ignition due to ignition from the process containment
30
failure/ sparks. The values suggested apply to both oil and gas releases. The origin of the data is not known.
See Table 1.32.
Table 1.32 Self-ignition probabilities
Leak rate (kg/s)
0.3
0.75
1.5
3
6
12
24
48
96
The values are derived from the correlation: p self ignition = 0.01 x m /100
Where m is the mass release rate in kg/s.
The values are low, and are unlikely to contribute significantly to the overall ignition probability. In any case,
the values may only be relevant to leak situations where a catastrophic failure occurs in the containment.
The DNV TDIM Project [15] also analysed ignition events to estimate the occurrence of self-ignition
(i.e. where the leak and ignition had a common cause/ dependency). This concluded that a common leak and
ignition effect may only be present in ~ 0.1 % of releases, giving a probability of ignition due to this mechanism
of approximately 0.001. This is in reasonable agreement with the values suggested in Table 1.32, and has been
adopted within the ignition model.
1.4.5
Recent developments, both onshore and offshore, have enabled methods to be proposed that are able to:
Address the build-up of gas in confined or congested spaces such as in offshore modules.
Allocate ignition probabilities based on the area of a zone that would be exposed to a flammable vapour
cloud.
The two ignition models proposed (DNV TDIM and WS Atkins - HSE) use scenario specific release and
ignition data to estimate the probability of ignition. Both models can be used to estimate the potential for
ignition within the immediate plant and beyond.
In addition, both models are intended for use with vapour cloud releases. However, only the DNV
TDIM method includes data for oil releases.
The WS Atkins methodology is also limited in that it ignores the ignition contribution from equipment
in zoned areas; however the method could be adapted to include ignition sources in this zone if required.
The DNV TDIM model also includes a correlation to assess the average gas concentration in the
module. However, this does not indicate how much of the module would be exposed to a flammable gas
concentration. The JIP Gas Build-up Workbook does assess the volume of the gas cloud within the module.
Where the release is able to form a free jet, then only the volume of this jet would be available to reach ignition
sources. The situation becomes more complex if the gas is shown to build-up such that the module becomes
saturated with gas at some equilibrium concentration profile. In these cases the model would give the fraction
of the volume containing vapour within the flammable range, but this would only relate to the equilibrium
condition. Other areas of the module may have seen flammable concentrations during the gas build up phase.
There is therefore a question of how many ignition sources could have seen a (brief) transient flammable gas
concentration during the gas build-up phase, and how many would be within the equilibrium zone after the
build-up has occurred. In practice, this may not be a major concern, since the Gas Build-up JIP Workbook
method typically gives fractions of the module within the flammable range at equilibrium of 60 % so any
ignition probability based on this cannot be more than a factor of two out, even if the entire module was subject
to gas in the flammable range at some time during the release. In practice, the brief exposure times during gas
31
build-up would discount many ignition mechanisms, e.g. from hot surfaces, and reduce the contribution from
intermittent ignition sources. So the error would be less than a factor of two.
1.5
1.5.1
Model approach
Two different approaches could be taken to the development of the model. One approach would be to represent
each potential release scenario separately, taking account of directional aspects and the specific locations of any
ignition sources. The overall probability of ignition could then be estimated by summing all the contributions
from the various scenarios and averaging the result. This appears to be the approach taken by the WS Atkins
model and DNV TDIM model.
An alternative approach would be to treat a range of related scenarios of a similar type and location as
a generic release and also use generic or averaged ignition source densities for the surrounding areas to derive
the averaged probability of ignition. In theory both approaches should give the same answer. However, the first
method would also be able to indicate the likely distributions of ignition probabilities for the scenario e.g. for
various wind directions and provide scenario specific values for a given wind direction for instance, whereas
the second method could only ever give the overall average value. The downside to the specific approach
would be the need to run the model many times to consider all the different release directions, wind direction,
etc.
It could be argued that the existing WS Atkins model and DNV TDIM model already provide a means
to assess the ignition probabilities of specific scenarios. The current need may be for a simpler model more
suited to use in QRAs which have to consider a large number of generically grouped scenarios.
1.5.2
The ignition data available from the literature and OIR12 data reviews do not provide a sufficient basis on
which to develop detailed ignition correlations that can take account of factors such as the ventilation rate,
module configuration, gas dispersion etc. The development of a simple empirical model is therefore considered
impractical at this stage.
The data and model review also show the relatively weak foundations of ignition correlations models
such as those presented by Cox, Lees and Ang[2], E&P Forum[6] etc. These are anchored to a few points, the
primary one being blowout data that, according to the analysis by Per Holand[5], is now out of date. There is
reasonable evidence from the Per Holand blowout data and the OIR12 data analysis that the average probability
of ignition of massive releases is of the order of 0.1 or less, not 0.3. The OIR12 data also indicates that an
average ignition probability of 0.02 to 0.04 is representative of small to medium sized gas and liquid leaks.
This suggests that a revised Cox, Lees and Ang equation for ignition probability would be: 0.01 at
<1 kg/s, 0.02 to 0.04 at 1-50 kg/s and 0.1 at >50 kg/s, with these data being equally applicable to both oil and
gas. The suggested revised values are shown in Table 1.33. Given the closeness of the revised figures to
those already in use for oil, it is suggested that the current oil ignition distribution could be retained for future
assessments, but that the gas distributions could be adjusted to the revised values for any future assessments.
Table 1.33 Generic ignition probabilities (Cox et al and Revised)
Release rate category
Minor
Major
Massive
Cox et al
gas leak
0.01
0.07
0.3
Cox et al
oil leak
0.01
0.03
0.08
Revised:
oil & gas
0.01
0.03
0.10
These revised values, or correlations based on these, could be adopted for scoping QRAs. However, a different
approach is needed if a more robust and differentiating ignition model is to be developed for use in detailed
QRA.
One alternative approach is to develop a model based on an area or volume basis using the ignition
source density approaches adopted in the DNV TDIM and WS Atkins methods. The DNV TDIM method can
provide generic data for use within zoned areas and the WS Atkins method has generic data for use in
unclassified areas and off site areas onshore.
32
The historical ignition data can be used as benchmarks to check that any ignition model developed
gives results in line with experience.
A distinction should be made between continuous and intermittent ignition sources. The contribution
of intermittent sources to the ignition probability will depend on the duration for which the ignition source is
exposed to a flammable concentration. In the immediate plant area, intermittent sources generally arise from
electrical equipment. These would typically be shutdown once a gas leak had been confirmed. Scoping
calculations on the time for gas jets to form and for an enclosed module with a typical 20 acph ventilation rate
to become filled with gas at LFL, indicate this will occur in a few seconds to a minute. This would be about the
time the leaks would be confirmed and equipment shutdown. It would therefore appear sensible to take the time
interval for intermittent sources as the expected shutdown time for electrical equipment.
1.5.3
Data on the timing of ignition and the resulting type of flammable event are sparse. The review in 1.2.1
indicates that for offshore blowouts, approximately 40 % ignite immediately (see 1.2.1.5), with almost 50 %
occurring over an hour after the blowout started. Onshore MAH data derived from an analysis of LPG
incidents in an accident database indicate 30 % of ignitions occurring within 30 s of the leak and 70 % within
2 min. (Table 1.11). By contrast, 95 % of the ignitions reported in the OIR12 data ignited immediately. This
may reflect the nature of many of the ignitions in the OIR12 data related to internal turbine or flare/vent
events or hot work activity rather than large process leaks spreading to meet ignition sources.
For QRA purposes, the common practice to assume a 50:50 or 30:70 distribution between early and
delayed ignitions appears to be reasonable in the light of the data available.
The timing of ignition is sometimes used as a means to determine the nature of the ignited event
resulting from ignition. An overview of ignition timing data is presented in Table 1.34. Early ignition is taken
to indicate a jet fire or pool fire depending on the material concerned (gas or liquid). Delayed ignition is taken
to indicate that the initial ignition would result in a flash fire or explosion. Some analyses assume a historically
based flash fire to explosion ratio to determine the relative frequency of these events. However, this does not
take account of the specific layout of the area in terms of its confinement and congestion, which affects the
chance of an explosion developing.
Table 1.34 Ignition timings overview
Plant type
Plant
Transport
Pipelines
CMPT blowouts
OIR12 offshore
1
0.22
0.53
0.24
0.10
0.94
> 1000
1.0
1.0
1.0
1.0
1.0
A review of offshore data from WOAD (Table 1.12) indicates a 21 % incidence of an explosion given ignition.
Cox, Lees and Ang (Table 1.3) have suggested a mass release rate based correlation for an explosion given
ignition ranging from 0.04 for small gas releases, 0.12 for medium releases, to 0.3 for massive releases.
The analysis of the OIR12 data (Table 1.26) indicates the relative likelihoods given in Table 1.35.
Table 1.35 Relative incidences of ignited events
Event type
Gas
Explosion
Flash fire
Jet fire
Liquid (oil/ condensate)
Explosion
Flash fire
Pool fire
14
58
28
7
53
40
33
These results support an overall distribution of early to delayed ignition of approximately 30:70 to 40:60,
early:delayed. The incidence of an explosion is less than that quoted for blowouts, but this may reflect the
relatively minor nature of many of the ignited events in the OIR12 database blowout and other releases may
be more likely to form a large cloud of gas in a confined/congested configuration capable of causing an
explosion.
It is therefore concluded that risk assessment approaches based on a 30:70 to 50:50 split in
early:delayed ignition or jet/pool fire:flash fire/explosion are reasonable.
The incidence of an explosion can be more problematic. A simple distribution such as that suggested
by Cox, Lees and Ang, or a simple fraction e.g. 21 % from blowout data, could be adopted for relatively generic
screening risk assessments. However, the detailed risk assessment of a particular facility may wish to review/
modify these simple factors depending on the nature of the leaks and plant configurations being assessed.
1.5.4
The more advanced ignition models being considered as the basis for a QRA ignition model use the vapour
cloud as the basis for the model. Separate calculations would be needed to estimate the vapour generated from
flashing releases, 2-phase releases or evaporating pools. In evaluating the model basis, consideration will need
to be given as to what release/ dispersion modelling should be included within the ignition model and what
should be excluded. For example, it may be better to include any assessment of the flash fraction in the general
process calculations in the QRA, rather than trying to apply a simplified model within the ignition model itself.
The Gas Build-up JIP Workbook[9] could provide a means to assess the volume of the plant area
exposed to flammable vapour concentrations. However, this method has shortcomings modelling small releases
where the momentum jet would dominate the dispersion. Here, a simple correlation based on the volume of a
free momentum jet may be more applicable. Although the workbook method indicates a maximum flammable
volume of approximately 50 % of the total volume, other areas of the module may see flammable gas during the
build-up phase. It may be best to assume that in the event of a module reaching this saturation condition, all
continuous sources of ignition within the module are considered, and the 50 % factor is only applied to the
intermittent ignition source densities, since only 50% of the volume would be within the flammable range for a
long period of time.
The workbook method also does not directly address areas with mechanical ventilation systems. If an
area has only mechanical ventilation, this may be used to calculate an effective ventilation rate, which could
then be used with the method to estimate the gas cloud size. Although the method is not intended for this
application, it should be able to offer a crude indication of the gas build-up.
Consideration will also need to be given on how to deal with spills of flammable liquids resulting in
pools or running streams, which could then ignite. Where these generate significant amounts of vapour, this
will dominate the ignition potential. However, if the material is of low volatility, and there is little potential for
aerosol generation (i.e. low pressure releases of low volatility liquids) - how should this be assessed? The DNV
TDIM model does include ignition source intensities for use with oil, presumably on a unit area basis. This
could be combined with a pool size correlation to estimate the ignition probability of a flammable liquid pool.
However, it should be noted that the size of a pool depends on many factors including the viscosity,
temperature, volatility, surface characteristics and time - it seems unlikely that a simple correlation could
successfully account for all these factors. In practice it would be the spray/ aerosol/ flashing components
generated from any spill that would dominate the ignition potential. The chance of ignition of a true liquid pool
or stream would be very low. A crude but simple alternative may be to assume a minimum pool depth and time
window to allow an assessment of the pool area.
Another approach would be to adopt a generic evaporation rate and use this to determine the
equilibrium pool size. Results for generic low volatility liquids (Class C) are presented in the IP15 supporting
calculations [13], indicating typical pool evaporation rates of 0.0006 kg/m2/s. However, this gives enormous
pool sizes (see Table 1.36) an estimate based on the pool size after 60 s may be more representative in terms
of assessing the ignition potential. Some switch would also be needed to deal with areas with grated floors or
specific bunding or drain arrangements.
34
Pool mass,
(kg), if 2 mm
deep
Volume released
after 60 s (m3)
277
2767
27667
69167
276667
3.6
36
361
904
3614
0.0072
0.072
0.72
1.8
7.2
Note: nominal 2 mm pool depth assumed for early stages of a hydrocarbon liquid spill. Actual value will
depend on which mechanism/ phase of pool spreading is dominant, the material properties and the surface
characteristics, etc.
1.5.5
Control factors
Several control steps can be taken to limit the chance of ignition, and it would be helpful if the more important
of these were considered within the ignition model.
Those controls thought to be most important relate to:
Isolation and blowdown of the process; limiting the inventory of flammable material available and reducing
the flow rate as the pressure falls.
Detection leading to the shutdown of electrical equipment and the stopping of hot work; reducing the
number of potential ignition sources.
Isolation and blowdown affects the gas dispersion characteristics. One way to address this would be to limit the
dispersion based on the available inventory.
Shutdown of hot work and electrical ignition sources could be addressed by only considering these
sources of ignition during the initial stages of the releases.
The operation of deluge systems is also sometimes used to help disperse releases and inhibit ignition.
There is some anecdotal evidence that the operation of a deluge system can cause ignition, perhaps due to static
generation, particles in the deluge stream, shorting of faulty electrical equipment or cooling effects on hot light
fittings leading to gas ingress. Overall it is not clear how effective deluge operation would be in terms of its
effect on the chance of ignition. This may be a topic for future work.
1.5.6
Historical data on the timing of ignitions and the relative incidences of fires and explosions have been reviewed.
The data are sparse but do cast doubt on a frequently used assumption that delay ignitions of vapour/ gas
releases result in explosions. Onshore plant data and the OIR12 data indicate a significant proportion of
immediate ignition (see Table 1.34). By comparison, WOAD indicates the chance of an explosion given
ignition of gas of 0.21 (see Table 1.12), and the OIR12 data gives 0.14 (see Table 1.35). It should be noted that
these are historical averages over a range of situations, so may not be appropriate to a scenario specific ignition
model. For onshore assessments, the timing of ignition can be estimated by considering the distribution of
ignition sources and estimating the time of travel of the vapour cloud to a specific ignition source. A similar
approach could be developed for a zoned ignition model.
The factors determining whether a vapour or 2-phase release ignites to give a jet fire, or a flash fire/
explosion followed by a jet fire, are complex. The ventilation regime and the timing and strength of ignition
will be important, but no method has been identified from the literature that could use these factors to assign
which of these outcomes would predominate.
Choices need to be made as to whether the probability of an explosion is based on some overall
historical averaged value, or on some scenario specific characteristics, or on a simple mass flow rate based
correlation such as that suggested by Cox, Lees and Ang. As with the release source modelling, consideration
needs to be given as to whether it would be better to include an approach to assign fire vs. explosion
probabilities within the ignition model, or whether this should be assigned outside the model. Our current view
is that this should be assigned outside the model. The nature of the release event and its environment, together
35
with the ignition timing distribution from the model can be used to assign suitable fire and/ or explosion
outcome probabilities.
1.5.7
If the calculations to distinguish between the liquid and vapour/ aerosol components of a release are to be
carried out within the ignition model, some means of classifying the releases will be needed. One option would
be to use the IP classification (Area classification code for petroleum installations (IP15), Annex A:
Classification and Categorisation of Petroleum Fluids, Draft document for review only) [17] as given in
Table 1.37 and Table 1.38.
Table 1.37 IP fluid categories
IP fluid
category
N/A
Suggested categories
Non-flammable
Liquid
Liquid
G
G
Vapour
Vapour
Definition
Liquids held at low /near ambient pressures or more than 5 C
below their flashpoint
Liquids held at pressure and more than 5 C below their
flashpoint but well below their normal boiling point - potential
mist formation
Liquids held above their flashpoint but well below their
normal boiling point
Any flammable liquids that could partially vaporise on release
(<40 % adiabatic vaporisation), i.e. held above their flash
point and close to their normal boiling point
Any flammable liquids that would substantially vaporise on
release (>40 % adiabatic vaporisation) i.e. above their
flashpoint and well above their normal boiling point
All flammable gases
All flammable 'liquids' held above their boiling point within
the process
Description
Whilst a number of issues are raised above about the scope and approach of the model, our view is that a simple
model based on typical plant area characteristics and suitable for use within QRA spreadsheets should be the
focus of this phase of the project.
Further, it is considered that any calculations of the flash fraction are carried out external to the model
as part of the QRA release source term calculations. This would leave the model to focus on releases that were
either gaseous (airborne) in nature and/ or forming a liquid pool.
It is suggested that the model be based on a four Area approach, using the area of each as a basis for
assigning an ignition probability (with a volume basis adopted for the immediate Plant Area).
36
The first three Areas are treated as concentric circles/ annuli characterised by radii equivalent to a
circular/annular representation of the actual area in the Area (see Figure 1.10).
Figure 1.10 Ignition modelling Areas
1-Plant Area
3-Site Area
4-Offsite Area
The ignition probability contribution in the first Area (the Plant Area) would be estimated for vapour releases
by the following steps:
1.
Assign the areas to enable the model to calculate the Area characteristic radii.
2.
Assign the module type in terms of its confinement and blockage (based on Gas Build-up JIP
Workbook factors f5 and f6 respectively for a range of typical configurations from open plant area to
confined module).
3.
Assign a typical windspeed for the Area (or the ventilation rate if mechanically ventilated?).
4.
Check the length of the momentum jet from the release (to LFL).
5.
Check the Gas Build-up JIP Workbook method to see if the Plant Area would become saturated at the
gas release rate specified, i.e. reach 50 % of the Plant Area volume.
6.
If the length of the jet is less than the characteristic radius of the Area (i.e. on average would not reach
a wall/edge) and the gas build-up is below saturation, then the volume of the flammable Area would
be based on the jet volume to LFL, otherwise, use the Gas Build-up JIP Workbook correlation to
assess the volume of the module.
7.
Assign a characteristic ignition source density based on a range of selected area types ranging from a
zoned plant with few sources to unzoned plant.
8.
Assign a detection/shutdown time to indicate the time when intermittent ignition sources should be
switched off (default 60 s).
9.
Assign a probability of hot work for the Area in terms of the number of hours per year when hot work
may be present.
37
10. Using the flammable volumes calculated and the ignition source densities assigned, calculate the
probability of non ignition for the Area.
The ignition probability contribution in Areas 2 to 4 would be assessed by considering the dispersion of vapour
from the Plant Area:
1.
Use the Gas Build-up JIP Workbook correlations to determine the ventilation rate in the Plant Area and
hence the concentration of gas leaving it.
2.
If the concentration is below LFL, assign an ignition probability of zero to all other Areas.
3.
If the concentration is above LFL, feed the concentration and mass flow rate into a correlation for the
dispersion of the vapour (select if based on D5 conditions for onshore or D10 for offshore or just use
D5?). This gives the hazard range - cloud length and crosswind distance. The area could be
approximated to a triangle or an ellipse.
4.
The dispersion correlation would be curtailed based on the available inventory where applicable (user
to specify if a limiting inventory is relevant and give the inventory).
5.
Use the cloud length to assess which Areas (2-4) can be reached by the cloud, and then use the
pertinent characteristic radii and cloud length/crosswidth to estimate the cloud areas in each Area.
6.
Assign ignition source characteristics to each of the Areas (including hot work hours and shutdown
time where appropriate) and then combine these with the cloud areas to calculate the probability of non
ignition for each Area.
The procedure for liquid releases would be similar, but would be based on the pool surface area in the various
Areas (these would be entered directly so the user can take account of the floor type etc?). The areas would be
combined with liquid ignition factors/intensities to calculate the ignition probabilities. Hot work and
shutdown timing factors would be used as for the vapour procedure.
The overall probability of ignition is then given by 1 minus the overall probability of non-ignition,
which in turn is the product of the probabilities of non ignition for all the four Areas.
P ign = (1-P no ign) where
P no ign = Area 1-4 (1-p ign in Area)
The model output would include the ignition probability overall and the probabilities for each Area. This could
help users to assess the impact of different ignition locations and risks if required.
1.6
MODEL DESCRIPTION
A model is proposed for the assessment of the ignition probability of hydrocarbon releases in QRA. The model
estimates the volume or area of flammable gas in a given plant area, and then combines this with suitable
ignition source densities to calculate the overall ignition probability. The model has been structured to consider
the ignition of hydrocarbons within the immediate plant area where the leak occurs, and any additional
probability of ignition should the flammable cloud spread to adjacent plant areas or beyond. This description is
based on the model as implemented in Version C1 following internal technical review and checking and
comments from the Joint Industry Steering Group.
1.6.1
Model structure
The model uses a four Area structure that represents the ignition source types, which could be reached by any
flammable gas/ vapour or liquid arising from the release.
38
1.6.1.1
Ignition model areas
The model considers the following Areas:
1.
2.
3.
4.
The first three Areas are treated as concentric circles/ annuli characterised by radii equivalent to a
circular/annular representation of the actual area in the Area (see Figure 1.11).
Figure 1.11 Ignition modelling Areas
1-Plant Area
3-Site Area
4-Offsite Area
The annular approach creates an implied uniform wind rose and takes account of the probabilistic area
interaction between any drifting flammable cloud or jet and the next plant area.
Another way of using the model is to treat the areas as a sequence of Areas, each characterised by a
dimension which is equivalent to the radius of an annulus. This approach could be used to look at specific
directional scenarios (Figure 1.12).
Figure 1.12 Alternative ignition modelling Areas approach
R3
R2
R1
X1
X2
X3
39
To evaluate this arrangement in the model as a direction specific scenario, the areas (m2) of Areas 1 to 3 would
need to be calculated and entered in to the model such that they give the Area 1 to 2, Area 2 to 3 and Area 3 to 4
interfaces at the relevant radii (R1 to R3) equivalent to the plant area dimensions X1 to X3 respectively.
The model distinguishes between gaseous and liquid releases. Liquid releases can be treated as
pressurised releases prone to mist/aerosol formation with rain-out to form a pool where appropriate, or as
simple non-volatile liquids forming a liquid pool. An overview of the model is presented in Figure 1.13.
Figure 1.13 Ignition model structure
Input
Data
Gas vs
Liquid?
Liquid
Rain-out
Correlation
Gas
Vapour / aerosol
Jet reaches edge of
Area
1 and exit at <LFL? Yes
No
Area 1 area based
on momentum jet size
Liquid contribution
- based on pool area
in Area 1 and 2
Gas leaving
Area above LFL?
Simple Pool
Area Correlation
Yes
No
Output
Data
Ignition
Source
Characteristics
1.6.1.2
Fluids considered
If the release is gaseous, then the release parameters are fed direct into the gas dispersion section of the model.
If a liquid release is input, the model estimates the degree of mist generation from this and treats this as a
gaseous release. Any rain-out is treated as a liquid pool. The model does not calculate the flashing of
superheated liquids or volatiles. However it does include some simple correlations for assessing the hazard
ranges associated with liquid jets of pressurised LPG and jets of Class B and C liquids. If the scenario is critical
it is suggested that evaporation and dispersion calculations should be performed external to the ignition model
and the relevant flammable areas for each model Area be entered in the model directly.
The mist/ rain-out step can be bypassed by setting the release pressure to zero; this forces the model to
treat liquid releases entirely as a liquid pool.
The model treats the liquid component as a pool and combines the pool area with a nominal vapour
layer height to give a flammable vapour volume. This is factored to give an equivalent area in terms of the
ignition model characteristics, and this is then combined with the relevant ignition source data (which are area
based) to give the liquid ignition contribution. The pool is allowed to spread within Area 1 and in to Area 2
and Area 3 if the pool area is greater than the Area 1 surface area. Any further spread beyond this is seen as
unlikely and is ignored.
The gas/ vapour source term is specified as a mass release rate. This is fed into either the Gas Build-up
JIP Workbook method for assessing the volume of flammable vapour in an enclosed area, or a free momentum
jet model correlation for an open area to determine the volume of flammable gas in the area (Area 1 - Plant
Area). This is then factored to give an equivalent area and combined with the relevant ignition source data
(which are area based) to give the gas ignition contribution for Area 1. The concentration of the gas at the
40
edge of Area 1 is then calculated, and if this is above the lower flammable limit (LFL) of the material, then the
model calculates the dispersion using either a passive dispersion model or a jet dispersion model (jet used in
openj (open jet) specific mode only - see 1.6.2). The model calculates the overlapping areas between the
dispersion flammable gas cloud and Areas 2 to 4, and combines these with the relevant ignition source data
(which are area based) to give the gas ignition contribution for Areas 2 to 4.
The model includes a dispersion limitation based on the inventory of material able to be released. This
includes checks/limitations on the volume of flammable gas in Area 1, and any subsequent dispersion into
Areas 2-4. The inventory limits for Areas 2 to 4 are set by taking the minimum of two dispersion ranges, one
for a continuous release at the specified mass release rate, and one for a quasi-instantaneous release of the total
inventory.
1.6.1.3
Ignition characteristics
The ignition modelling includes a hot work factor, which can be specified in terms of the hours of hot work per
year in the Plant Area. This ignition source is added to the more general sources in there. The model includes a
number of typical ignition area types with established ignition densities and characteristics taken from the
literature. Each Area needs to be assigned an ignition area type from this list - these characteristics are then
combined with any hot work characteristics to provide the Area based ignition probability. The ignition area
types include continuous and discrete/intermittent source characteristics. The model calculates an effective
contact time between the ignition sources and flammable hydrocarbons to use for the discrete source
calculation. This is based on the time for gas to build up in the Plant Area, the duration of the release (based on
a decaying inventory/ mass release rate) and the time to isolate any electrical equipment and stop hot work in
the Area.
The ignition contributions from gas and liquid components from all four Areas are then combined to
give the overall probability of ignition.
A crude indication of the ignition timing distribution is also calculated, based on the time for the gas or
liquid to build up in Area 1 and the time for any subsequent dispersion to reach Areas 2, 3 and 4.
1.6.2
Input data
The model input data consist of basic information on the release, the four Areas being considered and their
ignition characteristics, and a number of options for modelling the release and any subsequent dispersion
beyond the immediate Plant Area. The data input sheet is shown in Table 1.39.
Table 1.39 Ignition model input data
Input data
General data
Windspeed (m/s) (typical)
Blank
Leak Type, gas or liquid or LiquidA or LiquidC
Mass flow (kg/s)
Pressure (barg)
Calculated Liquid flow (kg/s)
Estimated Liquid Pool Area (m2)
Maximum Liquid Pool Area (m2), optional
manual input
Liquid Phase Density (kg/m3), optional
Vapour Mole Wt (kg/mol)
Calculated density at ntp
Gas/ Vapour Density (kg/m3), optional entry
Calculated gas mass flow (kg/s)
Material Inventory (kg)
Calculated gas/ aerosol Inventory (kg)
LFL of vapour, volume fraction
Notes/ Comments
41
Notes/Comments
set to zero if not relevant
Time to shut off ignition sources
length along/down module
width across front face of module
height of front face of module, min=1m
The yellow boxes are for data entry, the blue boxes are for optional inputs (some of which over-ride
parts of the model) and the red boxes are values calculated by the model - for information.
The calculated gas density (based on an ideal gas) can be over-ridden by inserting the known value in the
cell below.
The Maximum liquid pool area (m2) optional manual input can be used to manually specify the maximum
pool area where this may be restricted e.g. by a bund or grated flooring.
42
Specific pool or flammable gas regions can be specified for each Area if detailed dispersion or pool spread
modelling results are available externally to the model in this case the model can be used to calculate the
ignition probability based on these values only (i.e. all internal release and dispersion models are
bypassed).
1.6.2.1
Ignition characteristics
The characterisation of each Area's ignition sources is based on selecting one of the 17 types given in
Table 1.40; this lists the types currently in the model, together with summary descriptions and input entry
reference numbers.
Table 1.40 Ignition area types set up in model
Plant area Ignition area type
Description
Summary description
Offshore basic process module
1 Offshore process
2 Offshore compression
Offshore compression/ pump process module
Offshore power generation module
3 Offshore power generation
4 Zoned light process
Zoned light equipment process area
5 Zoned medium process
Zoned medium equipment process area
6 Zoned heavy process
Zoned heavy equipment process area
7 Unzoned light process
Unzoned light equipment process area
8 Unzoned medium process
Unzoned medium equipment process area
9 Unzoned heavy process
Unzoned heavy equipment process area
10 Storage
Outdoor storage area
11 Onsite offices
Onsite offices
12 Offsite industrial
Offsite industrial
13 Offsite urban
Offsite urban
14 Offsite rural
Offsite rural
15 Offsite vent
Offsite offshore cold vent and SBV
16 Offsite flare
Offsite offshore flare and SBV
17 No sources
Null no ignition sources
Note: SBV = Standby vessel
Input entry
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
In Table 1.40, zoned ignition source type categories refer to plant Areas in which area classification rules have
been applied (e.g. Zone 1 or Zone 2 area classification and where ignition source controls are in place).
'Unzoned' process area ignition characteristics are for Areas that are not subject to area classification or ignition
source control.
1.6.2.2
Plant configuration
The entry (see below) is used to input the configuration of the Plant Area (Area 1) and set how dispersion will
be modelled.
Fractional area of module front / rear
faces 'open', enter nosides if no walls,
open if open, openj if want to use jet
model for Area 2-4 dispersion
Entering a value between 0 and 1.0 means that the Plant Area is treated as a boxed tunnel with open ends each
of which have a fractional open area of the value entered; i.e. 1.0 is fully open, 0.5 is half open/ half enclosed.
The dispersion of vapour within the plant is modelled using the Gas Build-up JIP Workbook method.
Dispersion of any vapour leaving the Area is modelled using passive dispersion equations since any momentum
in the jet is likely to have been knocked out due to its interaction with equipment and boundaries in the Area.
Entering nosides causes the Gas Build-up JIP Workbook method to be applied but with the
confinement factors set to 1.0, i.e. no confinement, and the wind vectors factor to be set to 1.0, i.e. the wind can
blow directly into the Plant Area. This would be used to model a plant with a floor and ceiling but no side
walls. Dispersion of any vapour leaving the Area is modelled using passive dispersion equations since any
43
momentum in the jet is likely to have been knocked out due to its interaction with equipment and boundaries in
the Area.
Entering the open option causes the gas dispersion in the Area to be modelled as a free momentum
jet, but with passive dispersion from the edge of the Plant Area. This could be used for releases in very open
Plant Areas, but where equipment and buildings in adjacent Areas would limit any jet momentum, resulting in
passive dispersion beyond the immediate Plant Area. Dispersion of any vapour leaving the Area is modelled
using passive dispersion equations.
Finally, there is an openj option, which forces the model to treat both dispersion in the immediate
Plant Area, and all Areas beyond as a momentum jet release. This may be suited to modelling high pressure
releases in very open sites.
The Gas Build-up JIP Workbook method used for Area dispersion is not really validated for modelling
fully enclosed areas with forced ventilation systems. However, the model does include the option to enter a
ventilation rate for these situations, which is then used in the workbook method to calculate the flammable gas
volume and exit concentration. The dispersion of any gas leaving the Area above the LFL (i.e. from the exhaust
vents) would be modelled as a ground level source passive gas cloud.
1.6.2.3
Modelling complex geometries
The model can be applied to many different geometries, but this requires an understanding of the structure of
the model and how it works; some degree of judgement may be required to find the best way to model very
complex geometries. The model considers the following Areas:
1.
2.
3.
4.
The first three Areas are treated as concentric circles/ annuli characterised by radii equivalent to a
circular/annular representation of the actual area in the Area (see Figure 1.14). The fourth (outer) Area is
assumed to be infinite, extending from the outer edge of Area 3.
Figure 1.14 Ignition modelling Areas
1-Plant Area
3-Site Area
4-Offsite Area
The annular approach creates an implied uniform wind rose and takes account of the probabilistic area
interaction between any drifting flammable cloud or jet and the next plant area.
Although the model talks about this annular arrangement of Areas, in practice the model itself only
uses the annular radius when calculating the areas in each of Areas 2 to 4 affected by the vapour cloud. The
model simply assesses the areas of the trapezoids of the cloud within each Area, as defined by the cloud initial
width and its width at either the edge of the Area or where the cloud ends (whichever is shorter), and the
centreline distance from the near to the far edge of the Area (annual radius) or to the end of the cloud,
whichever is shorter. Figure 1.15 shows how the affected Areas (i.e., Areas 2-4) are calculated by the model.
44
R3
R2
R1
A2
A3
A4
X1
X2
X3
1.6.2.4
Modelling non-annular configurations
Another way of using the model is to treat the areas as a sequence of Areas, each characterised by a dimension,
which is equivalent to the radius of an annulus. This approach could be used to look at specific directional
scenarios (Figure 1.12).
Figure 1.16 Alternative ignition modelling Areas approach
R3
R2
R1
X1
X2
X3
To evaluate this arrangement in the model as a direction specific scenario, the areas (m2) of Areas 1 to 3 would
need to be calculated and entered into the model such that they give the Area 1 to 2, Area 2 to 3 and Area 3 to 4
interfaces at the relevant radii (R1 to R3) equivalent to the plant area dimensions X1 to X3 respectively.
This method would allow the model to be used to approximate specific plant geometries that are
different to the concentric circle configuration adopted within the model.
45
4
3
1
Could be modelled as (a) to represent the situation with the release exiting the left hand side, and the top and
bottom of the inner zone:
3
1
And as (b) below to model the situation with the release exiting the right hand side of the inner Area (with the
annular width of the blue annulus set to equal the width of the Area 2):
4
3
2
1
The overall ignition probability could then be estimated as [3 x (a) + (b)] / 4 if the direction is uniformly
distributed. Alternatively, the wind rose could be used to assign directional probabilities for the two cases when
calculating the average ignition probability for the release scenario.
More complex geometries could be handled in a similar manner.
46
C
E
S
This situation could be modelled as three separate situations, and then the results combined to give the overall
averaged ignition probability: The annular widths of A and C being set equal to the widths of A and C
respectively.
Release to the west (W), i.e. from B through A, with the width of annulus A set to equal the width of
module A.
Release to the east (E), i.e. from B through C, with the width of annulus C set to equal the width of module
C.
Releases to the north (N) and south (S), i.e. from B direct to Area 3, with the width of the annulus set to
equal the average width between Area 1 (B) and Area 4 (outer region beyond the yellow Area 3).
Wind
Gas
Wind
B
Plan View
47
Wind
Gas
Wind
B
Side View
1.6.2.5
Gas vs. liquid releases
The model has been designed to account for the potential for ignition from vapour/gaseous releases and liquid
pools. Liquid pools are treated as a surface with a flammable vapour layer above them (nominal 2 m deep
vapour layer). This is equated to a flammable vapour volume for the ignition calculation. Gaseous releases are
treated as gas clouds to determine a volume or surface area for the ignition calculation. The size of the gas
cloud is determined by the Gas Build-up JIP Workbook method, free momentum jet equations or passive
dispersion correlations depending on its location and plant configuration.
The model also includes a correlation to take account of mist formation and rain-out when a liquid is
released at pressure. It should be noted that the correlation is based on the release pressure and does not take
account of the hole size or other parameters which can affect the aerosol generation. It is considered that the
correlation is conservative in terms of the mist formed, and hence in terms of ignition potential. The mist/ rainout correlation is only activated if the pressure data input is set above zero (the pressure is not used for any
other purpose).
The model does not include an estimate of the flash fraction due to the release of superheated liquids.
Such calculations need to be made external to the model and the results fed into the model as a gas flow rate and
a liquid pool size where appropriate.
The model can handle a wide variety of situations by using the data entry appropriately, as described in
Table 1.41.
1.6.2.6
Inventory
The model allows the user to specify a limiting inventory for the release. This limits the dispersion of vapour
and liquid both within the release area, i.e. Area 1, and in the external to there, i.e. Areas 2-4. The liquid
inventory is used to limit the pool area. Within Area 1, the flammable vapour volume is limited to the volume,
which the inventory could form at the LFL. For gas dispersion in Area 2 to Area 4, the dispersion is limited to
the minimum of the continuous and quasi-instantaneous hazard ranges and cross widths.
Setting the inventory to zero (0) prompts the model to assume an unlimited inventory is available. The
continuous release model is used for dispersion in Areas 2 to 4.
When entering the inventory, the user should include both the actual locked in inventory of the system
which could be released plus any quantity of material that could be released prior to leak detection and plant
isolation, i.e. the inventory should include:
Total inventory able = (Initial release rate x detection and plant isolation time) + plant inventory
to be released
able to be released
48
49
1.6.3
Model output
The model output gives the overall probability of ignition, and a breakdown of the individual Area and gas/
liquid contributions (see Table 1.42). These contributions are only approximate since the ignition contributions
combine on a (1-pign) (i.e. product) basis. An approximate ignition timing distribution (cumulative
distribution) is also given. This provides the user with an indication of the timing and location of ignition.
Table 1.42 Model output data and format
Result, P ign =
Ignition contributions (approximate only)
Area 1 gas
Area 1 liquid
Area 2 gas
Area 2 liquid
Area 3
Area 4
Total
Ignition contributions (approximate only)
Gas
Liquid
Total
Ignition contributions (approximate only)
Area 1
Area 2
Area 3
Area 4
Total
Ignition probability vs. time, s
Time, s
Area
Area 1 early
Area 1 mid
Area 1 end
Area 2
Area 3
Area 4
Final Over all time
All
50
100.000 %
P ign
The model also provides information in a Summary Sheet (see Table 1.43).
Table 1.43 Model output information
Information from model:
Mass fraction of flow not raining out
Relative density ratio of vapour entering
Area 2
Flammable gas volume Area 1 (m3)
Flammable gas area Area 2 (m2)
Flammable gas area Area 3 (m2)
Flammable gas area Area 4 (m2)
Flammable pool area Area 1 (m2)
Flammable pool area Area 2 (m2)
Flammable pool area Area 3 (m2)
Pool area cross check (m2)
Vapour Cloud Hazard Range to LFL (m)
1.6.4
The ignition probability contribution in the first Area (the Plant Area) would be estimated for vapour releases
by the following steps:
1.
2.
3.
4.
5.
6.
7.
8.
9.
Assign the areas to enable the model to calculate the characteristic radii.
Assign the module type in terms of its confinement and blockage (based on Gas Build-up JIP Workbook
factors f5 and f6 respectively for a range of typical configurations from open plant area to confined module).
Assign a typical windspeed for the area (or the ventilation rate if a mechanically ventilated area - this rate
feeds directly into the Gas Build-up JIP Workbook method; see step 5).
In the open or openj options, calculate the length and volume of the momentum jet from the release (to
LFL). The volume should be limited to the volume of the jet within the Area (i.e. jet length no greater than
the effective Area 1 radius = (H L W / )^(1/2)).
For all other configurations, use the Gas Build-up JIP Workbook method to assess the plant volume in the
flammable range.
Assign a characteristic ignition source density based on a range of selected area types ranging from a
classified area zoned plant with few sources to unzoned plant. Adjust the area for the ignition calculation
based on the calculated flammable volume and the typical plant height contained within the model. Note
this height is not the actual plant height, but is used to relate the Plant Area volume to a surface area for the
ignition type characteristics. The surface area exposed to a flammable concentration is given by the volume
of the flammable cloud, divided by this typical plant height. This ensures the model reflects the changes in
ignition source densities for plants which are significantly shorter or taller than the nominal plant used to
set the ignition source densities.
Assign a contact time between the flammable vapour and any ignition sources based on the minimum (nonzero) time from: the time for the vapour to build up to a concentration above the upper flammable limit
(UFL) within Area 1; the duration of the release where inventory limited (based on a simple decay ratio);
the electrical/ hot work detection and shutdown time to indicate the time when most ignition sources should
be switched off.
Assign a probability of hot work for the area in terms of the number of hours per year when hot work may
be present.
Using the flammable volumes calculated and the ignition source densities assigned, calculate the
probability of non ignition for the Area.
The ignition probability contribution in Areas 2-4 would be assessed by considering the dispersion of vapour
from the Plant Area.
10. Use the Gas Build-up JIP Workbook correlations to determine the ventilation rate in the plant area and
hence the concentration of gas leaving the area. If the model is in open or openj mode, the Area 1
51
11.
12.
13.
14.
15.
dispersion is based on a free momentum jet. A jet-concentration correlation is then used to estimate the jet
concentration at the edge of Area 1.
If the concentration is below LFL, an ignition probability of zero is assigned to all other areas (i.e. the area
of the flammable cloud in these areas, Areas 2-4, is zero).
If the concentration is above LFL, the concentration and mass flow rate are fed into a correlation for the
dispersion of the vapour. This gives the hazard range - cloud length and crosswind distance. The area is
calculated using a simple area correlation derived from a 2 sigma (2 y) Gaussian cloud width.
The dispersion correlation is curtailed based on the available inventory where applicable. The minimum
(non zero) result of the continuous and quasi-instantaneous dispersion model hazard ranges is used as the
ignition model hazard range.
Use the cloud length to assess which Areas (2-4) can be reached by the cloud, and then use the pertinent
characteristic radii, initial cloud width and cloud length/ area correlations to estimate the cloud areas in
each Area.
Assign ignition source characteristics to each of the Areas (including hot work hours and shutdown time
where appropriate) and then combine these with the cloud areas to calculate the probability of non ignition
for each Area.
The procedure for liquid releases is similar, but is based on the pool surface area in Area 1 and Area 2 (if the
pool extends this far). The pool area is calculated assuming a pool depth of 2 mm and the liquid density
specified (default is 830 kg/m3), and is based on the size of the pool at the time at which the hot work/ electrical
isolation is achieved or the total inventory of liquid available to be spilt, whichever gives the smaller pool size.
Ignition probabilities are based on an adjusted area basis - the volume of the flammable area above the pool
divided by the standard plant height used in the model. This equates the vapour layer above the pool (assumed
as 2 m high based on a typical Zone 2 distance for Class I, II(2) and III(2) flammable liquid storage tanks in
IP15, Area classification code for petroleum installations, 1990 [18]) to a gas cloud and uses gas cloud ignition
data to estimate the ignition probability. Hot work and contact time/shutdown timing factors are used as for the
gas/ vapour ignition calculation.
The overall probability of ignition is then given by 1 minus the overall probability of non-ignition,
which in turn is the product of the probabilities of non ignition for all the four gas Areas and two liquid Areas.
P ign = (1-P no ign) where
P no ign = Area 1-4, gas and 1-2 liquid (1-p ign in Area)
In practice the model calculates natural logarithms of the probability of non-ignition and then sums these, and
takes (1-exponential of the sum) to give the overall ignition probability. This is because some of the ignition
correlations are in natural log format.
The model output includes the ignition probability overall and the probabilities for each Area. This
could help users to assess the impact of different ignition locations and risks if required.
An ignition timing distribution is also calculated based on the ignition probability at 0.001, 0.5 and 1.0
times the contact time for Area 1, and the time for the gas cloud to reach Areas 2 to 4, based on the distance and
windspeed (time = distance / windspeed).
Details of the various data and correlations used in the model are given in the following sections.
1.6.4.1
Ignition source characteristics
The following ignition source characteristics have been implemented in the model. These are intended to
provide characteristics relevant to most commonly occurring situations.
Onshore offsite land use
See Table 1.44.
52
Rural
(per hectare)
9.9 x 10-3
6.5 x 10-3
8.2 x 10-3
The data in Table 1.44 are converted to a per m2 basis for use in the ignition model (by dividing by 10,000).
Equivalent densities taken from Table D.2, Appendix D, A model for the ignition probability of
flammable gases, Phase 2, WS Atkins for HSE, Contract Research Report 203/1998, HSE Books [11]. Those
data are said to be based on the current HSE dispersion/ ignition model.
The ignition densities are considerably lower than those for the light industrial zones presented in the
following section. Type 12 Offsite industrial ignition area type, for example, give ignition probabilities
approximately 1/30th of those of Type 4 Zoned light process (though this does depend on the contact time
between the gas and the ignition sources). A key factor in this difference is that the ignition characteristics for
the process areas are based on flammable gases being present inside an enclosed plant area, whereas the offsite
characteristics are based on a gas cloud passing over industrial activity, much of which would be indoors.
Onshore on-site areas (not area classified)
See Table 1.45.
Table 1.45 Typical ignition source densities for onshore plant
Ignition sources
p
ta
ti
A
Heavy equipment
0.5
1
0.028
50
levels
Process areas
Medium equipment
0.25
1
0.035
50
levels
Process areas
Light equipment levels
0.1
1
0.056
50
Storage
Material handling
0.1
10
20
0.333
0.0333
10
Offices
Light equipment levels
0.05
1
0.056
20
Taken from Table 4.2, Development of a method for the determination of on-site ignition probabilities, WS
Atkins, HSE Research Report 226, HSE Books, ISBN 0 7176 2843 4, 2004 [12]
Land use type
Process areas
Note the characteristics in Table 1.45 based on areas within no area classification or specific ignition controls.
Where:
, is the ignition source density, sources per hectare
p, is the probability of ignition or ignition potential, the probability of ignition when the source is active and in
contact with a flammable gas
If the source is intermittent, ta defines the time the source is active and ti defines the time between each
activation (min.).
, the activation frequency (per min.) is given by, = 1 / (ta + ti)
a, the probability that the source is active is given by, a = ta / (ta + ti)
The probability of non-ignition at some time, t, is given by:
Ln Q(t) = A [ (1-ap)e-pt-1 ]
Where A is the surface area (hectares) of the zone subject to flammable gas concentrations (i.e. the flammable
vapour cloud area), and t is in the same time unit as the frequency (min.).
(From equations 3.4 and 3.5, page 7 of A model for the ignition probability of flammable gases, Phase
2, WS Atkins for HSE, Contract Research Report 203/1998, HSE Books.)
53
Hot work
For hot work, the values in Table 1.47 have been adopted.
Table 1.46 Hot work ignition characteristics
Hot work
p
0.5
ta
Hours hot work
in Area per year
ti
Year minus
hours of hot
work
A
= ta /
(ta + ti)
=1/
(ta + ti)
1 / floor area of
Area concerned
This represents hot work as a localised source of ignition in the area, present for ta hours per year, with a 0.5
probability of ignition if flammable vapour were to reach the hot work whilst it was ongoing. This is
representative of basic hot work in a plant area. It may be conservative for situations where hot work is
undertaken in tented/ positive pressure ventilated enclosures designed to prevent gas ingress.
Offshore plant - onshore plant with area classification
Table 1.47 DNV TDIM - Recommended ignition source densities
Source
Continuous
Unit
Discrete (per s)
Gas
Oil
Gas
Oil
3.30 x 10-6
5.70 x 10-5
3.40 x 10-8
5.80 x 10-7
Pump
Per item
6.40 x 10-5
4.40 x 10-3
1.20 x 10-7
3.40 x 10-6
Compressor
Per item
-3
1.50 x 10
1.50 x 10
-2
-6
2.70 x 10-5
Generator
Per item
3.50 x 10-3
3.50 x 10-2
6.20 x 10-6
6.20 x 10-5
-5
-4
-9
1.50 x 10-7
Other
Per m
1.40 x 10
4.10 x 10
2.70 x 10
8.40 x 10
Personnel
Per m2
5.50 x 10-6
2.60 x 10-4
2.80 x 10-8
5.80 x 10-7
(From Appendix K, Table 8.31 Recommended Ignition Source Intensities, Ignition modelling - Time dependent
ignition probability model, Det Norse Veritas, Report No 96-3629, Rev 04, February 1998) [15]
The data in Table 1.48 have been used to derive per area based ignition probabilities for use in the model for
three typical equipment area types.
Offshore basic process module (Type 1 Offshore process) - with no specific high ignition risk
equipment (e.g. pipes and vessels only). This is based on the per m2 electrical equipment, other and personnel
continuous ignition source densities and discrete densities times the relevant contact time, t (s) (i.e. the time the
sources are exposed to a flammable atmosphere).
Offshore compression/ pump process module (Type 2 Offshore compression) - taken as the basic
module plus three compressors and two pumps. The point source data for the compressors and pumps are
divided by the module floor area in the model to give a per m2 ignition source intensity.
Offshore power generation module (Type 3 Offshore power generation) - taken as the basic module
plus three power generators. The point source data for the generators are divided by the module floor area in
the model to give a per m2 ignition source intensity.
These data are intended for use in offshore processing areas, and could also be applied to onshore plant
areas which have Zone 1 or 2 area classifications.
The ignition probability is given by:
Probability
of ignition
Area of
module
exposed to a
flammable
vapour 2
cloud (m )
all continuous
x source
contributions
-2
(m )
all discrete
source
contributions
(m-2)
contact
time,
t (s)
The model takes specific account of the aerosol/mists formed by liquid releases and the vapour space above
liquid pools, i.e. the flammable vapour/mist from a liquid spill, using the gas ignition data. The oil ignition
data are not used.
These same ignition characteristics have been used together with a generic plant floor area to provide
ignition characteristics for electrically classified (zoned) process light, medium and heavy equipment areas.
54
These give ignition probability contributions of approximately 1/20th of the equivalent unzoned process area
ignition characteristics.
Offshore offsite ignition
The model includes ignition area types for two common offshore situations, to address releases that extend
beyond the main part of the platform. The cold vent situation assumes a SBV is present in the area, giving a
single continuous ignition source in the 500 m zone around the installation (vessels engine exhausts etc.). A
0.25 chance that the vessel is in the Area, and a 0.25 chance it is caught within the cloud (i.e. unable to move
out of the way when the release occurs), and a further 0.25 chance that the cloud is at a low enough elevation to
reach the vessel is assumed.
For installations with flares, a single continuous ignition source is assumed some 75 m from the centre
of the platform, with a nominal 0.1 chance that the gas could is sufficiently elevated, and directed to reach the
flare (approximately 1 in 6 random chance in vertical plane plus further reduction since most passive releases
likely to be relatively buoyant and so rise and not reach the flare). This is then added to the SBV contribution
as per the cold vent situation.
The cloud areas which could reach the SBV or flare are also checked in the model to ensure any cloud
beyond these sources is ignored for the ignition calculation, i.e. the maximum cloud areas entered into the
calculation are limited to 1/6th of the total circular area reaching the flare at 75 m radius and the SBV at 500 m
radius. The 1/6th ratio is based on the maximum area the cloud could take up within these zones assuming
pessimistically that the cloud width:downwind hazard range ratio is 1.0. This ensures that cloud regions beyond
these (which would not reach any more ignition sources), do not contribute to the ignition probability.
Although these are rather crude assumptions, the overall contribution of these sources to the ignition
probability is normally very small. However they can become significant at very high mass release rates
(>100 kg/s). Further work may be required to develop more realistic ignition characteristics for these areas.
1.6.4.2
Self-ignition
The model includes an allowance for situations where the cause of the leak could also cause the leak to ignite,
e.g. hot work cutting. This is a fixed ignition probability contribution value of 0.001 (0.1 %) taken from an
analysis of data in the DNV TDIM work [15].
The model does not take into account self-ignition of the material being released, e.g. due to the
material being above its autoignition temperature. In these situations suitable ignition probabilities should be
assigned based on the self-ignition characteristics, not this flammable zone model.
1.6.4.3
Ignition of liquids vs. gases
The ignition characteristics adopted for the model are based on the ignition of gases and vapours. Hydrocarbon
vapours exhibit low ignition energies, typically 0.2-0.5 mJ for common alkanes (Table 2.1, Ignition probability
of flammable gases, WS Atkins, HSE Contract Research Report 146/1997, HSE Books 1997 [10]). Methane is at
the higher end of this range. It is considered that the ignition characteristics adopted are directly relevant to
leaks of gases and vapours and vapour from the evaporation of liquid releases.
The same ignition characteristics data are also applied to mists/ aerosols generated from pressurised
liquid releases. This is based on the findings of a review by Shell Research (Combustion hazards posed by the
pressurise atomisation of high flash-point liquids, PJ Bowen and LC Shirvill, Shell Research Limited, J. Loss
Prev. Process Industries, 1994, Vol 7, No. 3 [19]). This concludes that atomised liquids with low droplet sizes
(<30 micron) can have minimum ignition energies of the same order, or even lower than that of a flammable
vapour. For example, aerosols of methanol and aerosols of n-heptane exhibit minimum ignition energies of
between 0.1-10 mJ over a range of droplet sizes from 10-50 micron. The paper also quotes work validated
against heavy fuel oil, diesel oil and gas oil, predicting minimum ignition energies of a few mJ for droplets of
30 micron size over a wide range of equivalence ratios (stoichiometric ratios).
Given the lack of knowledge and data on droplet formation and droplet ignition, the ignition model has
conservatively assumed that the gas released based ignition data can be applied to mists from pressurised
liquid releases.
1.6.4.4
Contact times
The model uses a contact time between the flammable vapour and the active ignition sources to determine the
contribution of discrete (intermittent) ignition sources to the ignition probability.
In Area 1, the contact time is based on the minimum (non-zero) of:
The Hot work/ Electrical shutdown time (as defined by the user input).
55
The release duration (where an inventory is specified), where time = 7 x gas inventory (kg) / gas release
rate (kg/s). Both these values are taken from the input summary sheet. The basis of the 7x factor is
discussed in 1.6.4.10.
The time for the gas to build up to above the UFL in Area 1 (where the release rate is sufficient to allow
this).
dC mass
dt
where
m& is the mass discharge rate from the release
V the enclosure volume
t=
ln
m&
If m& < Cvol gV the ventilation is sufficient to maintain the average concentration in the volume below Cvol .
Of course it may still be possible to exceed Cvol in parts of the volume.
The contact times in Areas 2 and 3 are based on the minimum non-zero value of the user specified
Area hot work / electrical isolation time and the release duration.
For Area 4, the contact time is based on the release duration, with a default of 30 min. where an
inventory is not specified.
For liquid releases resulting in vapour formation, the gas contact time, as defined above, is used to set
the contact time. Where the model is being used to represent simple liquid pool type releases (i.e. gas contact
time = 0), the contact time used is the user entered hot work/ electrical isolation time.
56
V/ L^3
0.1
0.01
0.0001
Workbook equation
0.001
0.01
0.1
0.001
m / p U L^2
The equation used to calculate the volume of gas in the flammable region V (m3) based on the workbook
method is:
V = minimum of [ 0.6 (L^3) ] and [ 150 (m/(p U))^3/2) ]
where:
m is the mass release rate (kg/s)
p is the gas density (kg/m3)
U is the ventilation velocity in the module (m/s) (this is a function of the external windspeed, a congestion
factor and a confinement factor)
L is the characteristic dimension of the plant volume (m), L=(length * width * height)^(1/3)
The internal ventilation velocity (U) is given by:
U = 0.3 Uw fo [1/(1+29 vb X / L) ]^0.5
where:
Uw is the windspeed (m/s)
fo is the fraction of module front/rear face open
X is the length of the module (m)
Vb is the volume blockage ratio in the module
A ventilation factor (f4) of 0.3 has been selected since this is typical of many of the possible orientations. This
value can be overridden by a manual entry if desired.
This assumes that the front and rear faces have the same fractional open area.
The flammable volume calculated by the workbook method relates to the steady state condition once
the release and ventilation regime has established itself. It includes a degree of conservatism as indicated by the
correlation line vs. the experimental data, as shown in Figure 1.17.
The volume of vapour in Area 1 is also limited by the available inventory; a check is made to ensure
the volume of the flammable vapour in the area does not exceed that theoretically possible based on the mass
inventory, vapour density and LFL concentration.
57
1.6.4.6
Liquid pool size
The size (area) of the liquid pool used in the model is based on a 2 mm deep pool at time when most sources of
ignition are shut down, with a default time of 5 min. (300 s) (where no shutdown time or inventory specified)
and with an inventory cut off for the pool size based on the liquid fraction of the inventory divided by the liquid
density (default taken as 830 kg/m3) and 2 mm pool depth.
The model uses the minimum of the liquid inventory based pool size and hot work/electrical
shutdown time x mass release rate based pool size where an inventory is specified. Where no inventory is
specified, the model uses the hot work/electrical shutdown time x the mass release rate based to calculate the
pool size. Where no hot work/electrical shutdown time is specified, the model assumes a 5 min. default
time.
The pool area is given by:
Pool area (m2) = mass liquid (kg) / [ density (kg/m3) x pool depth (mm) ]
Where the mass is the liquid inventory (kg) or liquid mass release rate (kg/s) x time (s)
1.6.4.7
Mist generation and rain-out
Pressurised liquid releases can generate mist/ aerosol leading to some fraction of the material becoming
airborne, and the rest raining out. The mechanisms for droplet formation are complex and can occur due to
stresses and turbulence at the release site and in subsequent impacts with objects in the path of the liquid jet.
Test data from cyclohexane trials at 350 K used to develop the RELEASE model indicate 50 % rain-out at
pressures as low as 2.1 bara, 33 % at 2.8 bara and 9.5 % at 3.9 bara respectively. The cyclohexane test results
were selected as the basis since this material is the best match to the type of hydrocarbons and conditions being
evaluated by the ignition model. Ref: Table A-7, page 118, RELEASE - A model with data to predict aerosol
rainout in accidental releases, D W Johnson and J L Woodward, CCPS, 1999, New York, ISBN 0-8169-0745-5
[14]
. The correlation developed from these data and used in the model to assess the fraction of liquid ending up
in the air is:
If Pressure, P (barg) <0.5, then Fraction in air = P*0.286/0.5,
else Fraction in air = 1-EXP(-(0.15*(1+P)^2)))
Note: this does not take account of any flash fraction due to superheating of the liquid. If there is a significant
flash fraction associated with the release, then separate calculations should be undertaken to estimate the flash
fraction. As a rule of thumb it is often assumed that the flash fraction would also entrain a similar quantity of
liquid as aerosol. Any significant flashing usually means that it can be assumed that all the material ends up in
the air as vapour or mist either from the original flashing or from subsequent evaporation of droplets/ films
formed on equipment. This mass flow could then be fed into the model as a gas type release. If any significant
liquid proportion remains as spill/ rain-out, this could be included in the ignition model by manually entering
the pool size from the liquid. See Figure 1.18.
The correlation is broadly in line with the test results presented in Calculations in support of IP15: The
area classification code for petroleum installations, IP, November 2001 [13] (See Table 1), in that releases above
10 bar are taken to fully entrain (no rain-out). However the IP data are based on the RELEASE model, which is
conservative in terms of the amount of rain-out. The correlation developed from the same experimental data for
this model is conservative in terms of the entrainment, since this is likely to dominate the ignition potential.
The correlation should only be taken as an approximation. Many other factors affect the degree of mist
formation including the hole size and shape. Further work may be needed in the future to find more realistic
means to estimate the degree of mist/ rain-out from pressurised liquid releases.
58
Fraction in Air
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
0
10
Pressure, barg
1.6.4.8
Gas momentum jet Hazard range and flammable volume
Correlations have been developed from theoretical jet equations. These are used to estimate the hazard range to
the LFL and the approximate volume of flammable gas in the jet, by equating the jet to a cone.
The following correlations have been adopted:
k M
x = f m& a
Cvol M g
1/ 2
Cvol
k M
= f m& a
x M g
1/ 2
59
Width
To determine an interaction area or volume of the jet with ignition source density it is necessary to specify a
width as a function of distance. Concentration fluctuations in the jet mean that ignition may occur outside the
mean LFL boundaries. The ignition region of a natural gas jet has been determined by Birch, Brown and
Dodson (Lees 2nd Edition 15/283 [20]). From Figure 15.142c (Lees 2nd Edition 15/283[20]) the potential ignition
volumes and areas for continuous ignition sources are estimated as:
V 0.02 L f
A 0.15 L f
3
2
where Lf is the distance to LFL. For intermittent ignition sources, the jet peripheries above are likely to lead to
overestimates of the ignition probability, since these may only exist for some small fraction of the time of the
release. The fluctuations above LFL would have a low probability of coinciding with the ignition source being
active.
1.6.4.9
Liquid jets
The hazard range from liquid jets (flashing liquids 2-phase jets) can be far greater than for the equivalent gas
jet due to the high momentum and evaporation in the jet. The hazard range is primarily a function of the release
diameter. However effects such as gravity, density, the degree of superheat etc. also play a part. Overall,
modelling of these releases can be complex, and it is suggested that in critical analysis a suitable 2-phase release
model is used to evaluate the range to LFL and resulting flammable vapour cloud area.
The model does include two simple correlations derived from the detailed modelling undertaken for
the IP15 (Institute of Petroleum Calculations in support of IP15: The area classification code for petroleum
installations, IP, November 2001, ISBN 0 85293 339 8 [13]). These are based on the results presented for fluid
categories A (LPGs), B (C68 to C10s with 20 % C3 to C5s) and C (C8 to C10s) liquid jets in Table B.5.
The correlations derived are:
For Fluid A:
60
Figure 1.19 Release maximum vs. average release rate Decay factor
1.2
0.8
0.6
0.4
0.2
0
0.1
10
100
1000
61
1/ 2
Q
L f = 12 0
u
ref
Q
L f = 5.4 0
u
ref
1/ 2
1/ 2
C0
C
f
C0
C
f
in open conditions
1/ 2
in urban conditions
The coefficients in the above have been adjusted to be appropriate for short time averaging, by multiplying the
15 min. average lateral spread by 0.5.
The boundary for passive plume widths is based on published correlations for the Gaussian standard
deviation e.g. Briggs based on about 15 min. averages
y 0.113x
y 0.226 x
in open conditions
in urban conditions
To account for fluctuations it is considered appropriate to go out to 2x the above standard deviations, giving for
area
2
62
A similar adjustment is made to the virtual source distance to ensure that the cloud width at the point of
exit from Area 1 does not exceed the effective width of the Area (taken as the geometric mean of the length and
width). The virtual distance is adjusted so that:
The distance to the virtual source from the edge of Area 1 Area 1 width / (2 x 0.2 in open
conditions and 0.5 in urban conditions).
U ref
If:
1/ 3
a Q0
g
<6
where W is the initial width of the cloud, Uref is the windspeed, and Qo the volumetric flow rate,
then dense or buoyant gas dispersion is more appropriate so long as the release is no longer momentum
dominated.
The model checks the dispersion source term against these criteria and indicates if the release is buoyant or
dense based on the molecular weight of the gas compared with air. The result is shown on the Summary Sheet
as a warning to the user. Buoyant releases are likely to give lower hazard ranges than those predicted by the
passive dispersion models implemented in the ignition model. Dense releases could go further than predicted
by the passive models.
1.6.5
The single scenario model that accompanied this report was issued as an MS Excel workbook at Version C1.
Note, however that this was superseded by Version D1 in Phase 2 and also issued in a multirun format. They
are simply referred to as the UKOOA Ignition model.
The workbooks contain the following key worksheets:
Read me a brief note about the model and its status plus a list of changes since Version A.2.
User guide a copy of Table 1.41 giving guidance on how to model different types of releases.
Results record records the results of any runs completed, copied from the Summary worksheet.
Summary the user interface, input/ output worksheet. This contains the liquid mist/ rain-out correlation to
calculate the gas vs. liquid mass release rates and pool sizes and the gas inventory. The output part of the
worksheet summarises the various ignition probability contributions to the overall probability and
63
calculates the percentage distributions of these contributions. Clicking the click to save summary
command button saves its contents to the Results record as formatted values. New records saved are placed
under any previous records.
Area 1 plant area: this contains the Gas Build-up JIP Workbook and momentum jet modelling
components and the ignition characteristics used for Area 1 and the Area 1 ignition vs. time distribution
calculation.
Areas 2-4: calculates the annuli radii and models the gas dispersion outside Area 1.
Area 2 ign to Area 4 ign: calculate the ignition contribution from these Areas based on the areas calculated
in worksheets Areas 2-4.
Area 1 liquid: contains the liquid ignition calculations for Area 1, including the time vs. ignition
probability contributions.
Area 2 liquid: contains the ignition calculations for any liquid overspill from Area 1.
Area 3 liquid: contains the ignition calculations for any liquid overspill from Area 2.
In addition, the multirun format workbook contains the following additional worksheets in support of its
capability to run the UKOOA Ignition model in a batchwise mode for a range of user defined scenarios and
release rates:
Data file: the user interface, input/output worksheet similar to the Summary worksheet but with the
capability to carry out batchwise calculations. Selecting the Save results to Results record saves the
contents of the Data file to the Results record, albeit without formatting. New records are placed under any
previous records.
Chart_all: displays the results of the ignition probabilities against flowrate for the runs studied in the Data
file.
All the ignition calculations are carried out using standard functions and cell formulae. The only visual basic
elements relate to the results save button and the workbook opening at the Read me worksheet.
1.7
MODEL BENCHMARKING
The model has been subjected to a detailed technical review and spreadsheet implementation checks as part of
its development.
The ignition model developed has also been run to provide estimates for a range of typical offshore
releases, and these have then been compared with typical ignition probabilities used in QRAs to-date.
The model (Version C1) has been run for a variety of release scenarios with an offshore processing
module as Area 1, with a compression module alongside (Area 2) and the rest of the installation (Area 3) treated
as (onsite) offices since this may be relevant to the accommodation unit. The ignition characteristics for
offices are also similar to those for zoned heavy process equipment, which could be present in other areas of the
installation. It is accepted that these Area 3 selections may be slightly conservative.
The results of these comparisons are shown in Figure 1.20 and Figure 1.21.
The UKOOA Ignition model estimates for low and high inventories in Figure 1.20 coincide for most
release rates, only diverging above 100 kg/s.
The comparison results show that the model is broadly in line with past data and correlations used for
assigning ignition probabilities. As expected, low volatility liquid pools (all liquid pool data) are estimated to
have a much lower chance of ignition than a liquid release that can generate a mist/aerosol. The levelling of the
curves at high flow rates are due to the areas becoming quickly saturated with gas/vapour. In some cases the
ignition probability is estimated to fall slightly at high release rates because the plant area reaches UFL more
rapidly and the inventory become exhausted more rapidly.
The ignition probabilities associated with high gas release rates in the tests are lower than those
commonly used. This is probably because the estimates used to-date in QRA for large releases are based on the
ignition of major blowouts from past experience largely in the 1970s and 1980s. In fact more recent blowout
incident analyses suggest a lower probability of ignition ~ 0.1 (see 1.2.1.5) is appropriate even for large
blowouts. Large releases from a processing area (as modelled in the tests) may also be less prone to ignition
than blowouts due to the fewer ignition sources and lower inventories involved.
The next steps should be to subject the model and its basis to a more comprehensive peer review and
case studies amongst safety practitioners, risk analysts and designers. This is to evaluate the technical validity
of the model and to assess its practicability, robustness and suitability for use in QRA.
64
Ignition Probability
1.00E -01
1.00E -02
1.00E -03
E&P Forum
- Gas
- Gas
- Gas
UKOOA Model
- low inventory
UKOOA model
- high inventory
1.00E -04
1.00E -05
0.1
10
100
1000
Ignition Probability
1 .00E -01
1 .00E -02
1 .00E -03
C o x, L e e s & A n g - O il
A E A T -H S E S tu d y - O il
C M P T - O il
1 .00E -04
U K O O A M od e l - o il m ist/ ra in o u t - lo w in ve n to ry
U K O O A M od e l - o il m ist/ ra in o u t - h igh in ve n to ry
U K O O A M od e l - a ll liqu id po o l
1 .00E -05
0.1
10
R e le a s e ra te , k g /s
65
1 00
10 00
1.8
A model has been developed that uses generic ignition source densities and dispersion models to provide an
estimate of the probability of ignition for hydrocarbon momentum driven releases. Dispersion modelling in the
immediate plant area utilises the Gas Build-up JIP Workbook correlations, or equates the release to a liquid or
gas momentum jet. Dispersion beyond the immediate plan area can be modelled as a passive gas cloud or a
momentum jet depending on the confinement or congestion in these areas. The model also provides the facility
to feed in the results from detailed dispersion modelling into the model as flammable cloud areas/ volumes and
pool areas for each Area. This overrides the internal dispersion models and provides a means to investigate
scenarios, which are outwith the models internal dispersion capability.
The model has been subjected to an internal technical evaluation review and limited benchmarking to
test its practicality and robustness. However, wider peer review and testing are needed before the model can be
accepted as a suitable method for use in QRA in support of safety cases and risk decision making. This is to
evaluate the technical validity of the model and to assess its practicability, robustness and suitability for use in
QRA. Therefore the next steps should be to subject the model and its basis to a more comprehensive peer
review and case studies amongst safety practitioners, risk analysts and designers.
It is also strongly recommended that the key cell formulae in the spreadsheet model are coded up (as
Visual Basic functions) to improve the clarity and ease checking and change control of the spreadsheet. It may
also be worth considering coding up the full model as a function or visual basic class object if it is to be taken
forward for use in QRA. This would allow the model to be used and run more easily and would also allow
improved quality and version control, and make future refinements and improvement easier to implement and
manage.
Once the model has been sufficiently validated, consideration should be given to running the ignition
model to assess the ignition probability of a number of typical release scenarios and plant configurations. The
results from these runs would form the basis of look-up tables, which could then be used as a means to select
appropriate ignition probabilities for use in QRA or for checking and review activities. This would provide a
simple and practical means to assign ignition probabilities in most situations, avoid the need to run the model
many times as part of every QRA, and allow the model to be used in situations where the QRA host programme
or method is not compatible with the spreadsheet version of the ignition model.
1.9
The following issues have also been identified during the Phase 1 work as areas where further work may be
required to improve the understanding of ignition and associated factors:
1.
The issue of ignition timing and the potential for vapour ignition to result initially in a jet fire, flash fire or
explosion. Could a correlation be developed to help indicate when each of these outcomes is the more
likely based on factors such as the release type, timing of ignition, dispersion factors, type of ignition etc?
2.
Potential for aerosol/ mist generation from liquid releases. A better understanding of this is required to
allow more appropriate correlations to be developed to support QRA generally as well as ignition
modelling. This should include related aspects of the flash fraction of multi-component streams and flash
fractions from superheated fluids. The current model uses a correlation, which may be conservative in
terms of the fraction of material that becomes entrained in the air as vapour or mist.
3.
Further data collection and analysis are needed to better validate the ignition source densities and
characteristics used by the model and to identify densities and characteristics for other situations relevant to
offshore and onshore QRA.
4.
Conduct further research into methods to allow the model to better represent open plant configurations such
as those on the upper decks of offshore installations and onshore plant such as tank farms, pipelines, valve
stations etc. The current model is largely based on the Gas Build-up JIP Workbook method, which is
suited to congested and confined situations such as largely enclosed offshore modules and onshore plant
between floor levels or with some walls/ weather shielding/ building confinement or self-shielding. The
assessment of open situations with the Gas Build-up JIP Workbook method model is sensitive to the height
of the unit selected, since this defines the area, dilution and mass flux of the resulting vapour cloud and its
dispersion. A simple momentum jet approach is an alternative included within the current model, but this
ignores any effect of congestion within the plant area.
66
10.
11.
12.
13.
14.
15.
16.
17.
18.
19.
20.
21.
Forsth (Det Norske Veritas), Technology Oil and Gas Journal (25 July 1983), pp 154-158
Classification of hazardous locations, Cox, Lees and Ang, Institution of Chemical Engineers (May 1991),
ISBN 0 85295 258 9
A guide to quantitative risk assessment for offshore installations, J Spouge, CMPT, Publication 99/100
(1999), ISBN 1 870553 365
A review of ignition probabilities for use in offshore installation quantified risk assessments, AEA
Technology, HSE OSD Research Project 3209 (May 1995)
Offshore blowouts, causes and trends, Per Holand, Norwegian Institute of Technology, Trondheim,
Norway (1996)
Hydrocarbon leak and ignition database, Report No 11.4/180, E&P Forum, (May 1992)
4th EGIG Report 1970-1998 Gas pipeline incidents, Doc. No. EGIG 99.R.0074, (December 1999)
HSE Contract Research Report 206/1999 Risk from gasoline pipelines in the United Kingdom, HSE Books,
ISBN 0 7176 1684 3 (1999)
Gas build up from high pressure natural gas releases in naturally ventilated offshore modules, JIP
workbook on gas accumulation in a confined and congested area, BP Amoco, CERC, BG Technology
(May 2000) (referred to as the Gas Build-up JIP Workbook)
Ignition probability of flammable gases, WS Atkins, HSE Contract Research Report 146/1997, HSE Books,
ISBN 0 7176 1431 X (1997)
A model for the ignition probability of flammable gases (Phase 2), WS Atkins, HSE Contract Research
Report 203/1998, HSE Books, ISBN 0 7176 1657 6 (1998)
Development of a method for the determination of on-site ignition probabilities, WS Atkins, HSE Research
Report 226, HSE Books, ISBN 0 7176 2843 4 (2004)
Calculations in support of IP15: The area classification code for petroleum installations, Institute of
Petroleum, ISBN 0 85293 339 8 (November 2001)
RELEASE A model with data to predict aerosol rainout in accidental releases, D Johnson and J Woodward,
CCPS, AIChE, ISBN 0-8169-0745-5 (1999)
Ignition modelling - Time dependent ignition probability model, Technical Report No. 96-3629, Rev 04,
DNV (February 1998) (referred to as TDIM)
Offshore hydrocarbon release statistics, 2001, HID Statistics Report HSR 2001 002, HSE (January 2002)
Area classification code for petroleum installations (IP15), Annex A: Classification and categorisation of
petroleum fluids, Draft document for review only, Institute of Petroleum [Superseded by IP Model code of
safe practice in the petroleum industry Part 15: Area classification code for installations handling
flammable fluids (IP15), Energy Institute, Third edition, ISBN 0 85293 418 1 (July 2005)]
Area classification code for petroleum installations, Institute of Petroleum, First edition, ISBN
0 471 92160 2 (1990) [Superseded by IP Model code of safe practice in the petroleum industry Part 15:
Area classification code for installations handling flammable fluids (IP15), Energy Institute, Third
edition, ISBN 0 85293 418 1 (July 2005)]
Combustion hazards posed by the pressurise atomisation of high flash-point liquids, PJ Bowen and LC
Shirvill, Shell Research Limited, J. Loss Prev. Process Industries (1994), Vol 7, No. 3
Loss prevention in the process industries, FP Lees, Butterworth-Heinemann, 2nd Edition, ISBN 0 7506
1547 8 (1996)
Workbook on the dispersion of dense gases, RE Britter and J McQuaid, HSE Contract Research Report 171988, HSE Books (1998)
67
68
SECTION 2
PHASE 2 REPORT LOOK-UP CORRELATIONS
2.1
INTRODUCTION
This Phase 2 project aims to develop simple look-up correlations for assigning ignition probabilities in
quantified risk analysis (QRA). The objective was to use the detailed UKOOA Ignition model developed in
Phase 1 to generate results for a number of selected representative scenarios, and to turn these results into
simple correlations that analysts could use thus avoiding the need to run the detailed model for the majority of
QRA applications, and also providing greater consistency. Initial work with the ignition probability model
during its development in Phase 1 showed that, for a given plant configuration and release substance, the
ignition probability could be estimated from a simple correlation based on the mass release rate. This has
formed the basis of the look-up correlations in Phase 2.
Section 2.2 provides a brief description of the UKOOA Ignition model: full details can be found in the
Phase 1 Report [1].
The representative scenarios and their configuration characteristics were selected and developed from a
mix of experience and actual plant data. These have been chosen to represent the majority of typical release
types addressed in onshore and offshore major accident hazard (MAH) QRA.
Modelling was then undertaken for all the selected scenarios. The results of this modelling are
presented in this report, which also gives full details of the basis of the scenarios modelled.
The results have then been used to develop a number of 'look-up' correlations for the range of
representative onshore and offshore QRA release scenarios.
The best-fit/ look-up correlations have been made available both in graphical form and as a simple
Microsoft Excel workbook cell formula based on the release mass flow rate. Details of the correlations are
provided in this report, should users wish to develop their own codes or spreadsheet functions for use in QRA.
The scenario modelling has also included selected sensitivity analysis to test the robustness of the
ignition probability results. This has shown that the model results for the scenarios modelled are reasonably
robust, with no cliff edge effects.
The Phase 2 report also includes a discussion on ignition timing and apportioning the probability
distribution for different end events to ignited scenarios, such as jet fires, flash fires, explosions, etc. This is
intended to provide guidance to safety practitioners and risk analysts when estimating ignition probabilities and
the relative occurrence of various different ignited outcomes in QRA.
2.2
MODEL DESCRIPTION
The UKOOA Ignition model assesses the probability of ignition of hydrocarbon releases for use in QRA by
combining established data and methods on gas build-up, gas dispersion, area ignition source characteristics,
etc. The model estimates the volume or area of flammable gas or liquid in a given plant area, and then
combines this with suitable ignition source densities to calculate the overall ignition probability. The model has
been structured to consider the ignition of hydrocarbons within the immediate plant area where the leak occurs,
and any additional probability of ignition were the flammable vapour cloud or liquid to spread to adjacent plant
areas or beyond.
2.2.1
Model structure
The model uses a four Area structure that represents the ignition source types, which could be reached by any
flammable gas/ vapour or liquid arising from the release.
The model considers the following Areas:
1.
2.
3.
4.
69
The first three Areas are treated as concentric circles/ annuli characterised by radii equivalent to a
circular/annular representation of the actual area in the Area (see Figure 2.1).
Figure 2.1 Ignition modelling Areas
1-Plant Area
3-Site Area
4-Offsite Area
The model distinguishes between gaseous and liquid releases. Liquid releases can be treated as pressurised
releases prone to mist/aerosol formation with rain-out to form a pool where appropriate, or as simple nonvolatile liquids forming a liquid pool. An overview of the model is presented in Figure 2.2.
Figure 2.2 Ignition model structure
Input
Data
Gas vs
Liquid?
Liquid
Rain-out
Correlation
Vapour / aerosol
Jet reaches edge of
No Area 1 and exit at <LFL? Yes
Gas
Liquid contribution
- based on pool area
in Area 1 and 2
Gas leaving
Area above LFL?
Simple Pool
Area Correlation
Yes
No
Output
Data
70
Ignition
Source
Characteristics
2.2.2
Input data
The model input data consists of basic information on the release, the four Areas being considered and their
ignition characteristics, and a number of options for modelling the release and any subsequent dispersion
beyond the immediate plant area. The data input sheet is shown in Table 2.1.
Table 2.1 Ignition model input data
Input data
General data
Windspeed (m/s) (typical)
Blank
Leak Type, gas or liquid or LiquidA or LiquidC
Mass flow (kg/s)
Pressure (barg)
Calculated Liquid flow (kg/s)
Estimated Liquid Pool Area (m2)
Maximum Liquid Pool Area (m2), optional
manual input
Liquid Phase Density (kg/m3), optional
Vapour Mole Wt (kg/mol)
Calculated density at ntp
Gas/ Vapour Density (kg/m3), optional entry
Calculated gas mass flow (kg/s)
Material Inventory (kg)
Calculated gas/ aerosol Inventory (kg)
LFL of vapour, volume fraction
Notes/ Comments
71
Input data
Hot work (hr. per yr.)
Hot work/ Electrical Shutdown Time (s)
Area 2 area (m2)
Area 2 Plant Type, enclosed, partial, or open
Manual Area 2 flammable gas area input (m2)
Manual Area 2 liquid pool area input (m2)
Area 3 Site area
Plant Area - ignition area type
Hot work (hr. per yr.)
Hot work/ Electrical Shutdown Time (s)
Area 3 area (m2)
Area 3 Plant Type, enclosed, partial, or open
Manual Area 3 flammable gas area input (m2)
Manual Area 3 liquid pool area input (m2)
Area 4 Offsite
Plant Area - ignition area type
Manual Area 4 flammable gas area input (m2)
Notes/ Comments
set to zero if not relevant
set to zero if not relevant
area of adjacent plant(s)
overrides all dispersion models
overrides all dispersion models
Partial Area 3 manual ratio (0.6 default)
set to zero if not relevant
set to zero if not relevant
total area of site
overrides all dispersion models
overrides all dispersion models
The yellow boxes are for data entry, the blue boxes are for optional inputs (some of which over-ride
parts of the model) and the red boxes are values calculated by the model for information.
The calculated gas density (based on an ideal gas) can be over-ridden by inserting the known value in the
cell below.
The Maximum liquid pool area (m2), optional manual input can be used to manually specify the maximum
pool area where this may be restricted, e.g. by a bund or grated flooring.
Specific pool or flammable gas regions can be specified for each Area if detailed dispersion or pool spread
modelling results are available externally to the model in this case the model can be used to calculate the
ignition probability based on these values only (i.e. all internal release and dispersion models are bypassed).
The characterisation of each Area's ignition sources is based on selecting one of the 17 types given in Table 2.2;
this describes the types currently in the model, together with their summary descriptions and input entry
reference numbers.
Full details of these ignition types can be found in the Phase 1 report [1].
For the ignition source types in Table 2.2, zoned refers to plant areas in which area classification
rules have been applied (e.g. Zone 1 or Zone 2 area classification and where ignition source controls are in
place). 'Unzoned' process area ignition characteristics are for areas that are not subject to area classification or
ignition source control.
72
2.3
Input entry
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
Phase 1 of this project [1] using the UKOOA Ignition model has shown that for a given scenario, the ignition
probability varies with the mass flow rate, and that this relationship can be represented by a relatively simple
correlation.
The objective of this Phase 2 project is to develop look-up tables or correlations for a range of
representative scenarios to provide an easy to use reference for ignition probabilities for use in QRA. These
tables/ correlations would be supported by guidance on how to select a suitable representative scenario, interpret
and apply the data, consider sensitivities etc. This is aimed at making the model easier to use. It avoids the
need to run the detailed model for every scenario, and uses the developer's detailed knowledge of the model to
provide the most appropriate means to represent the scenarios. The look-up tables should also ensure a more
consistent application of the UKOOA Ignition model in QRA.
The range of situations addressed by the look-up tables is intended to represent the majority of typical
QRA scenarios. This should mean that many QRAs would only need to use the look-up tables to select suitable
ignition probabilities for the scenarios being considered. If some of the scenarios being modelled do not fall
within the range of representative scenarios in the look-up tables, then the full model can be used to estimate an
appropriate ignition probability.
The UKOOA Ignition model (Version D1) has been used to model the representative offshore and
onshore scenarios to provide a series of simple mass release rate based ignition probability correlations for use
in QRA.
The following representative scenarios have been assessed as requested by the Joint Industry Steering
Group:
Release Types:
Gas releases;
LPG (flashing liquefied gas) releases, where appropriate;
Pressurised liquid oil releases leading to a spray release with flashing/ evaporation/ aerosol
formation;
Low pressure liquid oil releases leading to a spreading pool only (no aerosol formation or flashing).
Release Rates:
73
Onshore configurations:
Cross-country pipeline (high pressure natural gas, LPG, and a flammable liquid in either an industrial/
near urban area or in a rural area);
Small plant (high pressure vapour, 2-phase and flammable liquid spill for two typical small plant
configurations e.g. semi enclosed plant and an open but relatively congested plant on a small
manufacturing site);
Large plant (high pressure vapour, 2-phase and flammable liquid spill): this would involve the same
leak scenarios as for Small plant above, but with different plant and surrounding area dimensions and
ignition characteristics based on two large plant configurations e.g. within a refinery/ large
petrochemical complex.
Storage (multi-tank tank farm with LPG at pressure (medium and large inventories), and an ambient
flammable liquid e.g. petroleum spirit/ crude oil in two situations (medium and large inventories), one
within a processing plant complex and the other in a remote location).
Offshore configurations:
The detailed characteristics of these configurations are described in the following sections.
2.4
The modelling adopted a number of generic release materials (fluids) to be evaluated, and included a number of
general assumptions or bases that were applicable to all or main groups of scenarios. The general bases and
assumptions are detailed below.
2.4.1
Fluids considered
A range of release materials was considered to encompass the majority of offshore and onshore hydrocarbon
production and processing situations. These included gas (essentially taken as methane/ natural gas), LPG and
an oil based liquid with a small flash fraction, and one with no flashing.
A liquid flash fraction of 15 % was assumed for the offshore oil/ liquid scenarios (artificially fixed by
setting the release pressure = 0.27 barg) since offshore hydrocarbon liquids can contain significant light
fractions, as well as being subject to aerosol generation from edge effects from pressurised releases.
A liquid flash fraction of 10% was assumed for the onshore oil/ liquid scenarios (artificially fixed by
setting the release pressure = 0.175 barg) as it was considered that this may be more appropriate for the degree
of aerosol generation resulting from edge effects around the edges of the release with high pressure liquids, with
little or no flash fraction.
The gas/ LPG scenarios provide an indication of the ignition probability if all or most of the liquid was
expected to atomise or vaporise on release.
A summary of the fluid parameters and properties used in the model is given in Table 2.3.
74
Mole weight
of vapour/
flash
component
(kg/kmol)
17
UFL
(volume
fraction)
LFL
(volume
fraction)
Liquid
phase
density
(kg/m3)
0.15
0.05
LiquidA
50
0.10
0.02
500
100
0.10
0.02
850
LiquidC
100
0.10
0.02
850
Classification in
model
Gas
It should be noted that the ignition model includes a correlation to estimate the flash/ atomised fraction of liquid
releases based on the CCPS 'RELEASE' model work (See Phase 1 Report [1]); the main factor in determining the
fraction atomised is the release pressure. However, this correlation is based on test and analysis results for
small releases. The mechanical interactions causing aerosol formation may be largely due to edge effects
between the fluid and the hole edge/wall. Therefore, the correlation may not be suited to large releases through
large holes. This is one reason why the model scenarios selected for the Phase 2 project used predetermined
'flash' fractions more consistent with past QRA approaches and experience rather than the model in-built
correlation.
2.4.2
Release rates
A number of release rates were modelled, covering a wide range of spill sizes. These ranged from 0.1 to
1000 kg/s. Selected modelling at higher release rates was also undertaken where appropriate. The release rates
modelled are shown in Table 2.4; these were selected to provide suitable data points on a log-log plot, the most
appropriate form of presentation for the ignition model results.
Table 2.4 Release mass flow rates used
Release mass flow rate (kg/s)
0.1
0.3
0.5
0.7
1
3
5
7
10
30
50
70
100
300
500
700
1000
75
2.4.3
Windspeed
A windspeed of 5 m/s was selected as the basis for all the modelling. This is a typical onshore value, and an
analysis of UK North Sea offshore data indicated a median windspeed of approximately 5-6 m/s. Sensitivity
analysis at 2 m/s and 8 m/s shows no major differences in the ignition probability.
An average windspeed has been selected to provide look-up correlations reflecting averaged
conditions, since it is envisaged that the look-up correlation ignition probabilities will be used in QRAs
covering a wide range of weather situations.
If ignition probabilities are needed to reflect very specific wind conditions, then the full UKOOA
Ignition model should be used. This includes the ability to input manually the results from dispersion
simulations, overriding the simple dispersion correlations embedded in the model. This feature should be
utilised for specific dispersion conditions where detailed dispersion modelling results are available.
2.4.4
Pool sizes for all onshore spills (from plant, pipeline or storage) were generally taken to be limited to a
maximum of 100 m by 100 m (10,000 m2) by bunds, drainage systems, or other natural or man made
restrictions. Modelling was also carried out using a maximum pool size of 300 m by 300 m (90,000 m2) to take
account of very large crude oil storage tank areas and large multi-tank bunds. This was achieved by adding a
maximum pool size limit to the ignition model for these scenarios. No restriction was used for the offshore
scenarios since the overall installation sizes were smaller than 10,000 m2.
For onshore plant releases, an additional case was considered where the plant area could be considered
to be bunded (either by drains, bunds, kerbing etc), such that the spill would not spread out beyond the Plant
Area (Area 1). In these cases, the maximum pool area was limited to the smaller of 10,000 m2 and the area of
Area 1.
For pipeline releases, Area 3 (the total area of the site) was taken to be 10,000 m2 (i.e. approximating
to a 100 m pipeline corridor); this is the surface area limit to which any pool was allowed to spread.
The model assumes liquid pools cannot spread into Area 4 (i.e. offsite).
Version D1 of the UKOOA Ignition model includes the ability to specify a maximum (limiting) pool
area in the workbook Summary Sheet data input cell Maximum Liquid Pool Area (m2) optional manual
input (Cell B14). Alternatively, fixed liquid pool areas can be individually set further down the Summary
Sheet input data form for Area 1 through to Area 3; see Cells B41, B49, B57, respectively.
2.4.5
The model includes a specific ignition source type for offices on site. This includes a relatively high ignition
source density, and is intended for use in situations where flammable vapour clouds can enter an office
complex. The 'office' ignition source densities are used in the modelling to estimate the contribution to the
ignition probability for a process release due to the presence of offices or other accommodation areas.
Ignition source type '11 Onsite offices' has been used to represent accommodation units on offshore
installations since these typically consist of high-density living areas, with galleys, workshops, utility rooms etc.
The extent to which gas could enter these areas has been addressed by setting the area type to 'enclosed', so
there is limited ingress into the office area (up to 10% of the total area affected). This is considered
conservative given the typical design of offshore accommodation units to resist gas ingress, for example, by
maintaining a positive pressure in the area and having suitable heating, ventilation and air conditioning (HVAC)
inlets with gas detection and shut-off systems.
General (i.e. non-plant) areas of onshore sites have not been modelled using the office ignition type.
These areas typically consist of open ground, with access roads, car parking, various offices, stores, workshops
etc. Applying the office ignition type to mixed-use areas of this nature would be overly conservative. Type '12
Offsite industrial' provides a more suitable and appropriate ignition source for use in these areas; this has been
used for all onshore general site areas.
A summary of the area ignition types available in the model is presented in Table 2.2.
2.4.6
Selecting the 'open' ventilation configuration option for Area 1 (see UKOOA Ignition model Version D1
workbook, Summary Sheet Cell B35) would cause earlier versions of the model to return a #DIV/0 error if the
liquid modelled is being treated purely as a liquid, i.e. with no flashing or atomisation. This is because the
76
model was looking for a vapour source term from Area 1 into Area 2; since there is no vapour component, an
error was generated.
To avoid this error, the 'openj' ventilation configuration option (see Table 2.1) was selected for some of
the liquid (no flash or atomisation) scenarios instead of the 'open' option, even though the gas and flashing runs
had been modelled using the 'open' option due to the configuration of the area. This makes no difference to the
ignition modelling, since no vapour is involved, and prevents the #DIV/0 error in the workbook.
Version D1 of the ignition model includes an error trap to prevent this #DIV/0 error occurring when
a liquid only release is specified.
2.4.7
Pressures
Different values were used for the pressure of the material being released depending on the type of module or
system being evaluated.
It should be noted that the pressure has no effect on the ignition probability for gas or vapour releases
as calculated by the ignition model; the dispersion equations used for these are mass release rate based.
The pressure is used in the ignition model for liquid releases, and feeds into the vaporisation/ flash/
atomisation calculation, which determines the fraction of the mass flow that does not rain out or remain as
liquid. In general, the correlation used gives near total atomisation/ flashing above 5 barg (~99.5 % flash at
5 barg). Increasing the pressure above this would have little or no effect.
The pressure is also used in the ignition model to estimate the dispersion distances for flashing liquid
jets (LiquidA or LiquidC types); see Table 2.3. The dispersion distances for liquid jets are strongly correlated
to the hole size, so the pressure is used in-conjunction with the mass release rate to derive an effective hole size;
this is then used in the liquid jet dispersion correlation. This means that lower pressures will give longer hazard
ranges, since these imply a larger hole size for a given release rate. The effects are only likely to be significant
if the 'openj' configuration is selected (see Summary sheet Cell B35 in the UKOOA Ignition model Version D1
workbook) since, for these types of release, the model uses the liquid jet model to estimate the dispersion of the
vapour/aerosol in all Areas 1-4. The 'open' configuration would also use the liquid jet model in Area 1, but then
reverts to a passive dispersion model for Areas 2-4.
A flash fraction of 15 % was assumed for the offshore scenarios (artificially fixed by setting the release
pressure = 0.27 barg) since offshore hydrocarbon liquids can contain significant light fractions, as well as being
subject to aerosol generation from edge effects from pressurised releases.
A flash fraction of 10 % was assumed for the onshore scenarios (artificially fixed by setting the release
pressure = 0.175 barg) as it was considered that this may be more appropriate for the degree of aerosol
generation resulting from edge effects around the edges of the release with high pressure liquids, with little or
no flash fraction.
The pressures adopted for the representative scenarios are summarised in Table 2.5.
2.4.8
The Gas Build-up model [8] used in the UKOOA Ignition model [1] includes a factor (factor f4) to represent the
module geometry (number and location of any walls) and its orientation with respect to the wind direction. The
factor represents the component of the wind velocity driving the natural ventilation in the module (the module
internal air velocity). This is generally based on a cosine function of the wind angle relative to the module and
can vary between 0 and 1.0. The UKOOA Ignition model uses a default value of 0.3, which is conservative for
most tunnel or L-shaped configurations, and slightly optimistic for U-shaped configurations, depending on the
prevailing wind direction.
The layout of plant buildings and areas onshore is not usually wind direction dependent, so a range of
orientations is possible. There is also considerable variation in enclosed and semi-enclosed plant design. For
these reasons, the default value of 0.3 has been used for all congested/confined onshore plant scenarios, as this
should be conservative for most orientations and geometries.
The nature of offshore plant and equipment means that these are often aligned to be downwind of
accommodation areas, and as a result they tend to be placed in a particular wind alignment. A typical North Sea
wind rose has been used to estimate wind rose averaged configuration (f4) factors for a variety of offshore
module types. The results are shown in Table 2.6. This shows that typical tunnel type configurations give
average f4 values of approximately 0.6. L-shaped modules average nearer 0.5. FPSOs are a special case in that
they weather vane, so the wind is always blowing in more or less the same direction. This can result in f4
factors of under 0.2 for modules with transverse walls to the bow and stern sides. The specific module
orientation f4 factors in Table 2.6 have been used for the offshore scenario ignition probability modelling.
77
Onshore plant
Onshore storage
Offshore process,
wellhead or
compression module,
or riser, including
NUIs and FPSOs (i.e.
all offshore
scenarios)
Fluid
Pressure
(barg)
Gas
LiquidA
0
70 & 10
LiquidC
0.175 & 0
Gas
LiquidA
0
10
LiquidC
0.175 & 0
LiquidA
LiquidC
0.175 & 0
Gas
LiquidA
0-50
5
LiquidC
Basis
Pressure not used for gas releases
70 barg is typical high pressure LPG pipeline pressure;
sensitivity test run at 10 barg since this would give larger
liquid jet size
0.175 barg gives 10 % flash/ atomisation.
0 barg used for pure liquid releases, no flashing
Pressure not used for gas releases
10 barg used as lower end of typical LPG plant pressure
range, since this would give larger liquid jet size
0.175 barg gives 10 % flash/ atomisation.
0 barg used for pure liquid releases, no flashing
5 barg used as lower end of typical LPG storage pressure
range, since this would give essentially full flashing and
larger liquid jet size
0.175 barg gives 10 % flash/ atomisation.
0 barg used for pure liquid releases, no flashing
Makes no difference pressure not used for gas releases
5 barg used as lower end of typical condensate/ lights
processing pressure range, since this would give
essentially full flashing and larger liquid jet size
25 barg gives full flash of liquid typical separator
pressure
0.27 barg gives 15 % flash/ atomisation.
0 barg used for pure liquid releases, no flashing
Although there can be some variation in windspeed distributions between different sectors of the North
Sea, the directional probabilities are broadly similar, and the f4 factors estimated here should be suitable for
general QRA ignition modelling purposes throughout the North Sea.
2.4.9
The ignition model allows Area 2 and Area 3 plant types to be modelled as open, partially enclosed or enclosed
(see Summary Sheet Cells B47 and B55 in the UKOOA Ignition model Version D1 workbook).
The model uses these classifications to determine the maximum area of that area which can be affected
by flammable gas or vapour. The default values are 1.0, 0.6 and 0.1 respectively. The enclosed classification is
used for offices, administration buildings, accommodation units and other essentially enclosed spaces. The
partial classification is used for utility and process areas where there are some walls and wind wall/ louvres
restricting ventilation, but the unit is not totally enclosed. The partial classification has also been used for more
open process areas where the installation geometry is such that gas or vapour leaving Area 1 would not easily
pass into the adjacent Area 2 or Area 3, for example in the case of tunnel shaped offshore modules separated by
solid walls.
The 'partial' fractional area default of 0.6 is considered conservative for momentum driven releases,
since the Gas Build-up JIP Workbook [8] indicates that the extent of the flammable region within a module does
not exceed 0.6 (60 %) regardless of the ventilation configuration. The 'partial' classification is therefore
considered conservative for most offshore module scenarios, where releases will tend to be momentum driven,
and the 'side by side' geometry means that wind shear forces rather then simple downwind drifting will
determine the movement of gas between modules.
78
North
Straight through
f4 factor
1
0.707
0.1
0.707
1
0.707
0.1
0.707
0.5
0.646
0.6
Prevailing wind
79
L-shaped
f4 factor
1
1
1
0.707
0.23
0.163
0.1
0.707
0.5
0.543
0.5
U-shaped
f4 factor
0.1
0.707
1
0.707
0.1
0.1
0.1
0.1
0.25
0.292
0.25
The Gas Build-up JIP Workbook method is not appropriate to passive dispersion some way downwind of the
source, e.g. for large onshore facilities where the leak source could be surrounded by other plant areas. In these
cases, the flammable region could extend over Area 2 or Area 3. Even in these configurations, it is unlikely that
more than 60 % of the footprint area of Area 2 or Area 3 would be exposed to a vapour cloud within its
flammable region due to the nature of the plant layout and the directional nature of wind or momentum driven
dispersion.
2.4.10
The storage area ignition parameters used in Version C2 of the UKOOA Ignition model (and all previous
versions) were based on the work by WS Atkins for the Health and Safety Executive (HSE) [7]. This quoted
ignition parameters for a 'Storage External' area, which were adopted for the model. However, it has been
found that these ignition parameters can give overly conservative results, with leaks in storage areas having a
similar or greater chance of ignition than in process plant areas.
A more detailed review of the Atkins report shows that the 'Storage External' characteristics were
intended to represent storage areas (with or without area classification or ignition source controls) where there
was considerable fork-lift truck activity, e.g. a drum handling area or similar. The report indicates that open
storage areas on plant, which are area classified, i.e. have ignition source controls, should use the 'Classified
external' area ignition parameters (see Table 2.7). However, even these are based on frequent fork-lift truck
activity, and so they would not be representative of most onshore plant hydrocarbon tank farms. The fork-lift
truck activity is assumed to present a source for 5 min. in every 30 min. (giving a time interval of 25 min.).
Process plant tank farms would not typically have such frequent vehicle movements. However, there
could be some road tanker or rail tanker movements in or around the storage area, or other activity that may
present a (low) risk of ignition.
The 'Storage' ignition parameters used in the modelling for the look-up tables, has therefore, adopted
the 'Classified External' area ignition parameters, with an adjusted time interval of 115 min. (i.e. potential
sources in the area for five minutes in every two hours or for an average of one hour in 24 hours) (see Table
2.7). This adjustment is intended to reflect the lower activity in a typical tank farm compared. These revised
ignition parameters give 'Storage' leak ignition probabilities of approximately one tenth those of an equivalent
leak in a process plant area, one fifth of those for a leak in an offsite industrial area, and five times that for an
offsite rural area: these are more in line with expectations given the ignition characteristics of these areas. The
overall effect for typical storage release scenarios is to reduce the probability of ignition for liquid spills by a
factor of five to ten, compared to the previous Version C2 of the model. Large vapour spills from storage tanks
are not affected to the same extent, since these can extend beyond the storage area into areas with different
ignition characteristics.
Table 2.7 Storage ignition parameters
Land-use type
Ignition
potential, p
Storage external (Atkins [7])
Classified external (Atkins [7])
Storage (for this study) (Note 1)
0.1
0.05
0.05
Ignition parameters
Time source
Time
active, Ta (min.) interval, Ti
(min.)
10
20
5
25
5
115
Ignition source
density, (per ha)
10
10
10
Note 1: Version D1 of the UKOOA Ignition model includes these revised ignition characteristics for ignition
source type '10 Storage'.
2.5
The basis of the input data for the representative scenarios' Areas and ignition source characteristics are detailed
in the following sections. The relevant input data values are shown in Table 2.8.
2.5.1
Four pipeline configurations have been modelled: pipelines in an industrial area, and pipelines in a rural area
each with 'openj' or 'open' configurations.
80
The openj type treats the dispersion in all areas as an unbounded momentum jet and would be representative
of a release above ground.
The open type uses the momentum jet dispersion model inside Area 1 but then reverts to a passive
dispersion model for Areas 2-4: this would be representative of a release from a buried pipeline, e.g. one where
the jet interacts with the ground or other obstacles.
Onshore pipelines have been approximated to a small Area 1 area, 10 m long by 5 m wide by 5 m high.
These dimensions are not critical since the releases have been modelled as open or openj dispersion types,
which assume a jet in Area 1, so the Gas Build-up JIP Workbook method [8] is not used, since it is not relevant
to this situation.
The Area 1 area has been assigned ignition source type 4 Zoned light process, which is considered
conservative, but representative of say a pipeline valve station. The small size of Area 1 means that this will not
contribute significantly to the overall ignition probability.
Since cross-country pipelines are not normally sited adjacent to plant, Area 2 has been set to zero, i.e.
zero area and a null ignition source type 17 No sources.
Area 3 has been used to represent an approximately 100 m wide pipeline corridor (areas set at
10,000 m2), with ignition source type 12 Offsite industrial for pipelines in an industrial environment, and
type 14 Offsite rural' for those in a rural area.
The same ignition source types have been used for the Area 4 (Offsite Area).
The electrical isolation time for all three areas was set to '0', a null value that causes the model to use a
default time of 30 min. (1800 s) for any ignition sources to be active. This means that any liquid spills would be
allowed to spread for up to 30 min. at the specified mass release rate up to a maximum area of 10,000 m2 for the
purposes of assessing the ignition probability (see 2.4.4).
2.5.2
A small onshore plant has been modelled as a 25 m by 25 m square plant, with a height of 12 m. A volume
blockage ratio of 0.06 has been used, which is towards the bottom end of the range considered in the Gas Buildup JIP Workbook trials and development [8]: it is considered that most onshore plant would not be as congested
as a typical offshore module.
No allowance for hot working has been included, but this is only a minor contributor to the ignition
probability.
Electrical isolation has been taken to occur after 2 min. (120 s), a typical detection and shutdown time
for onshore facilities.
Two configurations have been considered:
A confined plant (tight) represented as a tunnel configuration with 80 % free area at either end. This is
intended to be representative of a semi-confined plant. This configuration uses the Gas Build-up JIP
Workbook method [8] to assess the flammable volume in the area.
An 'open' configuration, representative of a plant without walls or ceiling where momentum driven
dispersion would dominate in the plant, but this would then interact with structures and equipment to leave
a passive dispersion source entering Area 2 (Adjacent Plant and beyond). This configuration uses a
momentum jet model to estimate the flammable volume in the area.
The Area 1 area ignition characteristics have been represented as ignition source type 5 Zoned medium
process plant.
The Adjacent Plant has been represented as an open plant area of 1000 m2 (approximately 35 m by
30 m size) with 5 min. to achieve electrical isolation and ignition source type 5 Zoned medium process' plant.
The overall site (Area 3) has been taken to be 20,000 m2 in area (approximately 200 m by 100 m size),
with ignition source type 12 Offsite industrial.
The Area 4 offsite area has been assessed for both rural and industrial settings, using ignition source
types 12 Offsite industrial and 14 Offsite rural, respectively.
Liquid spills have been assessed both as: bunded with the pool area limited to the Area 1 plant area;
or, unbunded with the pool able to spread to a maximum of 10,000 m2.
2.5.3
A large onshore plant has been modelled as a 50 m by 50 m square plant, with a height of 25 m. Sensitivity test
runs using a height of 15 m were also undertaken, since some plants may not include very tall columns: these
showed little effect, with the 15 m ignition probability results being within 10 % of the 25 m results. A volume
blockage ratio of 0.06 has been used: this is towards the bottom end of the range considered in the Gas Build-up
81
JIP Workbook trials and development [8]. It is considered that most onshore plant would not be as congested as
a typical offshore module.
No allowance for hot working has been included, but this is only a minor contributor to the ignition
probability.
Electrical isolation has been taken to occur after 2 min. (120 s), a typical detection and shutdown time
for onshore facilities.
Two configurations have been considered:
A confined plant (tight) represented as a tunnel configuration with 80 % free area at either end. This is
intended to be representative of a semi-confined plant. This configuration uses the Gas Build-up JIP
Workbook method [8] to assess the flammable volume in the area.
An 'open' configuration, representative of a plant without walls or ceiling where momentum driven
dispersion would dominate in the plant, but this would then interact with structures and equipment to leave
a passive dispersion source entering Area 2 (Adjacent Plant and beyond). This configuration uses a
momentum jet model to estimate the flammable volume in the area.
The Area 1 area ignition characteristics have been represented as ignition source type 5 Zoned medium
process plant.
The Adjacent Plant has been represented as an open plant area of 2000 m2 (approximately 50 m by
40 m size) with 5 min. to achieve electrical isolation, and ignition source type 5 Zoned medium process
plant.
The overall site (Area 3) has been taken to be 100,000 m2 in area (approximately 500 m by 200 m
size), with ignition source type 12 Offsite industrial.
The Area 4 offsite area has been assessed for both rural and industrial settings, using ignition source
types 12 Offsite industrial and 14 Offsite rural, respectively.
Liquid spills have been assessed both as: bunded - with the pool area limited to the Area 1 plant area;
or, unbunded - with the pool able to spread to a maximum of 10,000 m2.
2.5.4
Onshore storage
An onshore storage tank farm has been modelled as two scenarios, a 100 m by 100 m LPG tank farm and a
larger 300 m by 300 m oil/ liquid storage tank farm, with a height of 20 m. A volume blockage ratio of 0.06
can be seen in the data entry sheet, but this is not used since all modelling uses the open or openj configurations,
which assume momentum jet dispersion within the storage area.
No allowance for hot working has been included, but this is only a minor contributor to the ignition
probability.
Electrical isolation has been taken to occur after 5 min. (300 s), a typical detection and shutdown time
for onshore storage facilities.
Two configurations have been considered:
An 'open' configuration, representative of a plant without walls or ceiling where momentum driven
dispersion would dominate in the plant, but this would then interact with structures and equipment to leave
a passive dispersion source entering Area 2 (Adjacent Plant and beyond). This would be representative of a
storage area within a plant or industrial estate.
An 'openj' configuration where dispersion in all areas is momentum jet driven. This would be more
representative of a very open storage area, well segregated from any plant or other buildings.
The Area 1 area ignition characteristics have been represented as ignition source type 10 Storage.
The surrounding plant (Area 2) has been represented as an open plant area of 500,000 m2
(approximately 500 m by 1000 m size) with 5 min. to achieve electrical isolation, and ignition source type 5
Zoned medium process. This would be representative of a storage facility within a large petrochemical or
refinery complex.
A stand-alone or separated storage configuration (e.g. a tank farm on a site but not within plant areas)
has also been considered: in this case, the Area 2 has been set to zero area, i.e. no surrounding plant.
For the storage area within plant, the overall site has been taken to be 600,000 m2 in area
(approximately 600 m by 1000 m size) for the LPG case and 700,000 m2 in area (approximately 700 m by
1000 m size) for the oil/ liquid case, with an ignition source type 12 Offsite industrial.
For the stand-alone/ separated storage situations, the overall site has been taken to be 20,000 m2 in area
(approximately 200 m by 100 m size) for the LPG case and 150,000 m2 in area (approximately 500 m by 300 m
size) for the oil/ liquid case, with an ignition source type 12 Offsite industrial, representative of the site
82
administration, workshops and general areas. An electrical shutdown time of 5 min. has been assumed for these
areas.
As mentioned in 2.4.4, the pool size was limited to a 100x100 m area for most tanks, with a
300x300 m case being used for very large oil storage tanks and large multi-tank bunds.
The Area 4 offsite area has been assessed for both rural and industrial settings, using ignition source
types 12 Offsite industrial and 14 Offsite rural, respectively.
2.5.5
Offshore NUI
An offshore NUI process module has been represented as a two-deck unit 27 m long by 21 m wide and 16 m
high.
The ignition characteristic for Area 1 has been set as type 1 Offshore process. The electrical
shutdown time has been taken to be 1 min. (60 s), typical for an offshore facility.
Dispersion in the Area 1 area has been modelled using the Gas Build-up JIP Workbook [8], with a
volume blockage ratio of 0.09 (average value between workbook extremes). An f4 geometry factor of 0.7
(see 2.4.8) and the 'no sides' configuration has also been adopted: these are based on a module with no walls,
typical of the open nature of many NUIs.
The Adjacent Plant Area has been taken to be utilities etc., with an area of 300 m2 (20 m by 15 m), a
1.5 min. electrical shutdown time, and ignition source type 4 Zoned light process. The Area 2 area has been
taken to be partially enclosed, limiting the area affected by any gas ingress.
The overall installation (Area 3) has been assumed to include a small control room/ temporary refuge
(TR) type office area (approximately 20 m by 5 m size): modelled as an overall installation area of 972 m2,
partially enclosed with type 11 Onsite offices ignition characteristics. An electrical shutdown time of 5 min.
has been assumed for the accommodation areas.
The Area 4 offsite area has been taken as that for open sea with a cold vent, i.e. type 15 Offsite
vent.
2.5.6
Offshore FPSO
An FPSO installation process module has been represented as 79 m long by 45 m wide and 15 m high. FPSOs
typically have large open decks for the process modules, which can be very large, over 150 m long. The FPSO
also weather vanes, so the wind will always tend to be in a similar direction, blowing down the deck. This
means that releases of gas could potentially reach a very large area. The area of deck that could be affected in a
gas leak will depend, amongst other factors, on where the leak originates. Leaks at the upwind end could affect
the entire length of the deck, whereas those at the down wind end would only affect a short section of deck. For
the purposes of ignition probability modelling, an average location has been assumed half way along the deck.
This means the leak can at most reach half the deck area. This is considered a reasonable approach to assess
average indicative ignition probabilities for use in QRA, where a wide range of release locations and directions
could arise.
It should be noted that the scenario selected represents leaks on the FPSO process decks, and not leaks
in the turret area or within internal storage tanks. Turret areas in particular can vary considerably in design:
specific ignition probability modelling or assessment should be undertaken for the situation concerned.
The ignition characteristic for Area 1 has been set as type 1 Offshore process'. The electrical
shutdown time has been taken to be 1 min. (60 s), typical for an offshore facility.
Dispersion in the Area 1 area has been modelled using the Gas Build-up JIP Workbook, with a volume
blockage ratio of 0.09 (average value between gas build-up trials and workbook development extremes [8]).
Appropriate f4 geometry factors of 0.7 (open module) or 0.175 (module behind transverse wall) have been used
(see 2.4.8) and the 'no sides' configuration has also been adopted: these are based on a module with no side
walls, typical of the open nature of many FPSOs.
The Adjacent Plant Area has been taken to be utilities etc., with an area of 1500 m2 (33 m by 45 m), a
1.5 min. electrical shutdown time, and ignition source type 4 Zoned light process. The Area 2 area has been
taken to be open; this will be conservative, since some areas may be partially or fully enclosed.
The overall installation Area 3 includes the accommodation areas (approximately 26m by 45m size),
modelled as an overall installation area of 6225m2, with enclosed offices with type 11 Onsite offices ignition
characteristics. An electrical shutdown time of 5 min. has been assumed for the accommodation areas.
The overall installation area used has been adjusted to allow for the 'process Area 1' area only being
half the full process deck area, i.e. 6225 m2 = 26x45 (accommodation) + 33x45 (utilities) + 79x45 ( process).
The Offsite Area (Area 4) has been taken as that for open sea with a flare, i.e. type 16 Offsite flare.
83
2.5.7
An offshore integrated deck platform wellhead module has been represented as a two-storey unit 26 m long by
18 m wide and 17 m high.
The ignition characteristic for Area 1 has been set as type 1 Offshore process. The electrical
shutdown time has been taken to be 1 min. (60 s), typical for an offshore facility. An allowance has been
included for 15 hr. hot work per year in the area.
Dispersion in the Area 1 area has been modelled using the Gas Build-up JIP Workbook, with a volume
blockage ratio of 0.06 (lower end value from gas build-up trials and workbook development [8] to reflect a
wellhead/conductor area). An f4 geometry factor of 0.6 (see 2.4.8) and a confinement factor of 0.4 have also
been adopted to reflect a tunnel type module with 60 % wind wall at either end.
The Adjacent Plant Area has been taken to be process modules, with an area of 940 m2 (36 m by
26 m), a 1.5 min. electrical shutdown time, and ignition source type 1 Offshore process. The Area 2 area
has been taken to be partially enclosed, reflecting the configuration such that only a proportion of the adjacent
area is likely to be affected by any drifting vapour from Area 1.
The overall installation Area 3 has been assumed to include the accommodation areas and utilities
areas as office area (approximately 46 m by 30 m size), modelled as an overall installation area of 2800 m2
(approximately 92 m by 30 m size), with enclosed offices ignition characteristics (type 11 Onsite offices).
An electrical shutdown time of 5 min. has been assumed for the accommodation areas.
The Area 4 offsite area has been taken as that for open sea with a flare, i.e. type 16 Offsite flare.
This will be conservative compared to installations only having a cold vent.
2.5.8
Offshore integrated deck platform process and compression modules have been represented as an Area 1 unit
30 m long by 20 m wide and 8 m high.
Additional modelling was also undertaken for a larger module of 35 m long by 30 m wide by 10 m
high.
The ignition characteristics for Area 1 have been set as type 1 Offshore process' or type 2
Offshore compression', as appropriate. The electrical shutdown time has been taken to be 1 min. (60 s), typical
for an offshore facility. An allowance has been included for 15 hr. hot work per year in the area; however, this
has minimal effect on the ignition probabilities (see 2.6.9.1).
Dispersion in the Area 1 area has been modelled using the Gas Build-up JIP Workbook, with a volume
blockage ratio of 0.09 (averaged value from gas build-up trials and workbook development [8] to reflect process
area congestion). An f4 geometry factor of 0.6 (see 2.4.8) and a confinement factor of 0.4 have also been
adopted to reflect a tunnel type module with 60 % wind wall at either end.
The process and compression modules have also been assessed for a more congested and confined
configuration, using a volume blockage ratio of 0.14 (upper value from Gas Build-up JIP Workbook
development and trials [8] to reflect a very congested area). An f4 geometry factor of 0.6 (see 2.4.8) and a
confinement factor of 0.2 have also been adopted to reflect a tunnel type module with 80 % wind wall at either
end.
Selected cases have also been assessed based on a fully enclosed process module with mechanical
ventilation, with rates of 15 and 30 air changes per hour (acph).
The Adjacent Plant Area has been taken to be process or wellhead modules, with an area of 900 m2
(30 m by 30 m), a 1.5 min. electrical shutdown time, and an ignition source type of 1 Offshore process. The
Area 2 area has been taken to be partially enclosed, reflecting the configuration such that only a proportion of
the adjacent area is likely to be affected by any drifting vapour from Area 1.
The overall installation Area 3 has been assumed to include the accommodation areas and utilities
areas as office area (approximately 46 m by 30 m size), modelled as an overall installation area of 2800 m2
(approximately 92 m by 30 m size), with enclosed offices type 11 Onsite offices ignition characteristics. An
electrical shutdown time of 5 min. has been assumed for the accommodation areas.
The Area 4 offsite area has been taken as that for open sea with a flare, i.e. type 16 Offsite flare.
This will be conservative compared to installations only having a cold vent.
2.5.9
Offshore riser
Ignition of an offshore riser release in the air gap has been represented as a Area 1 unit 30 m long by 30 m wide
and 6 m high: equating to a riser towards the centre of a large offshore platform, and as a unit 5m long by 5 m
wide and 6 m high: equating to a riser towards the edge of a large offshore platform or the middle of a small
84
riser platform. A sensitivity case has also been run with an Area 1 height of 12m, i.e. a large air gap but this
has little effect since the dispersion in the area is largely momentum jet driven; the Gas Build-up JIP Workbook
method [8], which uses the plant height, is not appropriate.
The ignition characteristic for Area 1 has been set as type 16 Offsite flare. The electrical shutdown
time has been taken to be zero, since this has little relevance in the air gap region.
Dispersion in the Area 1 area has been modelled using the 'open' and 'openj' configurations:
The openj type treats the dispersion in all areas as an unbounded momentum jet and would be
representative of a riser momentum jet release that did not interact significantly with the sea surface, walls
or solid floors/ decks.
The open type uses the momentum jet dispersion model inside Area 1 but then reverts to a passive
dispersion model for Areas 2-4: this would be representative of a release from a riser where the jet interacts
with the sea, floors, walls or other obstacles such that it then disperses as a passive cloud.
In the close confines of an offshore installation, the dispersion of a major riser release is likely to be largely
momentum jet driven, even where some interaction with surfaces occurs.
The Adjacent Plant Area has been taken to be process or wellhead modules, with an area of 900 m2
(30 m by 30 m), a 1.5 min. electrical shutdown time, and an ignition source type 1 Offshore process. The
Area 2 area has been taken to be partially enclosed, reflecting the configuration such that only a proportion of
the adjacent area is likely to be affected by any drifting vapour from the riser in Area 1.
The overall installation Area 3 has been assumed to include the accommodation areas and utilities
areas as office area (approximately 60 m by 30 m size), modelled as an overall installation area of 2800 m2
(approximately 92 m by 30 m size), with enclosed offices type 11 Onsite offices ignition characteristics. An
electrical shutdown time of 5 min. (300 s) has been assumed for the accommodation areas.
The Area 4 offsite area has been taken as that for open sea with a flare, i.e. type 16 Offsite flare.
This will be conservative compared to installations only having a cold vent.
2.5.10
This scenario is intended to represent very large releases of gas, vapour or aerosol with rain-out that can engulf
the entire installation, such as those from a major gas or oil blowout or riser rupture.
It is based on the basic installation configurations used for the offshore process module scenario
(see 2.5.8). These values have been adjusted to represent a release covering a large open deck (Area 1) and
reaching into lower partially enclosed lower decks, i.e. a process deck and cellar deck (Area 2) and the
accommodation/ utilities areas at two deck levels (Area 3).
Area 1 is treated as an open deck of dimensions 30 m long by 50 m wide, using the 'nosides'
configuration. The ignition characteristics are based on type 1 Offshore process since an open deck should
have few ignition sources.
Area 2 is treated as a partially enclosed process deck 50 m x 30 m and a cellar deck of 30 m x 30 m.
The ignition characteristics are based on type 2 Offshore compression', which reflects a mix of compression,
process and power generation types at these deck levels.
Area 3 adds an enclosed accommodation and utilities area of 45 m x 30 m on two levels (total area of
6600 m2). The Area 4 offsite area has been taken as that for open sea with a flare, i.e. type 16 Offsite flare.
Releases were modelled as gas, oil with 15 % flash fraction (pressure set at 0.27 barg) and oil with a
50 % flash fraction (pressure set at 1.1 barg).
The configuration is also reasonably representative of a major riser release engulfing the installation,
with Area 1 being similar to an open cellar deck/ air gap area, and with Area 2 being adjacent but above rather
than below.
85
0.8 + open
0
0
10
5
5
4 Zoned
light process
0.06
Open + openj
Module ventilation
configuration
f4 geometry/ orientation
factor, optional
(default 0.3)
Mechanical ventilation rate
(acph)
Open
0
300
0.06
17 No
sources
0
0
Open
0
120
Onshore
large plant
Open + openj
0.06
100
100
20
10 Storage
Offshore
NUI
0.7
Nosides
0.09
27
21
16
0.7 + 0.175
if wall
Nosides
0.09
0
300
86
Partial
300
0
90
Open
1500
0
90
Partial
940
15
90
0.6
0.4
0.06
26
18
17
15
60
Offshore
wellhead
module
78.8
45
15
1 Offshore process
0
60
Offshore
FPSO
Open + openj
0.06
300
300
20
10 Storage
0
300
Open
2000
0
300
Onshore oil
storage
5
5
Area 1 Plant Area (source of leak)
0
0
0
300
300
60
Onshore LPG
storage
0 (defaults to 0.3)
0.8 + open
0.06
25
50
25
50
12
25
5 Zoned medium process
0
120
Windspeed (m/s)
Onshore
small plant
Onshore
pipeline
(crosscountry)
5
Partial
900
15
90
1 Offshore process
15 + 30 (for
mechanical vent)
30/ 35 large
20/ 30 large
8/ 10 large
1 Offshore
process, 2
Offshore
compression
0.09 + 0.14 if
confined
0.4 + 0.2 if
confined
0.6
15
60
Offshore process or
compression
module
Partial
900
15
90
Open +
openj
0.6
0.09
30 + 5
30 + 5
6 + 12
16
Offsite
flare
0
60
Offshore
riser
Partial
2400
2 Offshore
compression
15
90
0.6
Nosides
0.09
30
50
8
1 Offshore
process
15
60
Offshore total
engulfment
Onshore LPG
storage
600000 (on
plant) 20000
(storage only)
Open
0
300
Open
100000
0
300
12 Offsite industrial
Onshore
large plant
Area 4 Offsite
700000 (on
plant) 150000
(storage only)
Open
0
300
15
Offsite
vent
Partial
972
0
300
Offshore
NUI
Onshore oil
storage
Enclosed
6225
0
300
Offshore
FPSO
Enclosed
2800
0
300
Offshore
wellhead
module
16 Offsite flare
Enclosed
2800
0
300
11 Onsite offices
Offshore process or
compression
module
Enclosed
2800
0
300
Offshore
riser
87
Note: A '0' in the manual/optional data entry boxes indicates a null value, i.e. no manual input; the word 'zero' is needed to manually enter '0' as the value.
Open
20000
10000
Open
0
300
12 Offsite
industrial, 14
Offsite rural
0
0
Onshore
small plant
Onshore
pipeline
(crosscountry)
Enclosed
6600
0
300
Offshore total
engulfment
2.6
2.6.1
Results presentation
The modelling results for the representative scenarios are presented in Figure 2.3 through to Figure 2.23 in
Annex 2.B. The mass release rate based ignition probabilities for gas and oil proposed by Cox, Lees and Ang[2],
which are often used in current QRAs, are also presented in Figure 2.24 for comparison.
2.6.2
Onshore pipelines
The results for the onshore pipelines show that large releases in industrial areas can reach very high ignition
probabilities, up to 1.0. Rural releases are less likely to ignite, only reaching 0.1 for LPG and less for gas or oil.
LPG releases give the highest ignition probabilities due to the longer liquid jet hazard ranges and lower
flammability limits (LFLs) compared with gas. The 'openj' configuration results give slightly lower ignition
probabilities than the 'open' configuration results for gas and LPG releases, since the open configuration uses a
passive dispersion model, which is likely to give larger cloud sizes in Areas 2-4. For 10 % flashing oil releases,
the 'openj' configuration gives slightly higher ignition probabilities at high release rates, probably due to the
longer hazard ranges of liquid jets compared to passive dispersion at lower release rates.
2.6.3
Onshore plant
The results for onshore small and large plant show that massive releases of gas or LPG can reach high ignition
probabilities (>0.5). Releases on plants in an industrial environment also show significantly higher ignition
probabilities than those in a rural environment though, as expected, this effect is more pronounced for small
plant than large plant.
2.6.4
Onshore storage
The tank farm results show that large LPG releases are estimated to have high ignition probabilities, reaching
1.0 in industrial areas.
The large releases of LPG from tank farms are estimated to have higher ignition probabilities than
equivalent plant releases. This is because releases within plant areas experience some dilution based on the
plant dispersion modelling for the Gas Build-up JIP Workbook [8]; releases from a tank farm experience less
initial dilution and so can travel further downwind increasing the probability of ignition as the cloud extends
over industrial areas. In addition, the storage area scenarios have higher inventories than plant scenarios,
allowing larger clouds to develop and spread further downwind.
Large LPG releases from a tank farm in a rural area reach an ignition probability of 0.1 at 1000kg/s,
although at higher release rates this does level off at approximately 0.5.
The ignition probabilities for liquid spills from tanks, with no flashing, are predicted to reach 0.015 at
high release rates. This is based on the assumption that the spill will be constrained by a bund or drains, roads,
kerbs, or slopes to a maximum size of 10,000 m2. A scenario has also been run to represent a large multi-tank
bund of 300 m by 300 m square (90,000 m2): this gives an ignition probability of up to 0.15 for very large
releases.
There are only slight differences between the 'open' and 'openj' configuration results for the tank farms;
the differences are greatest for large LPG releases, where the liquid jet option gives a slightly higher ignition
probability.
2.6.5
Offshore NUIs
The results for offshore modules follow the familiar S-curve of the Gas Build-up JIP Workbook [8]. The
ignition probabilities for releases on NUIs are predicted to be slightly lower than conventional staffed platforms,
due to their more open layout and smaller size.
2.6.6
Offshore FPSOs
FPSOs are predicted to have relatively high ignition probabilities due to the large size of these vessels and the
unfavourable alignment with the prevailing wind, but these are still lower than the maximums given by Cox,
Lees and Ang [2] (see Figure 2.24). Process modules on FPSOs located behind a transverse wall exhibit poor
dispersion even of small releases, giving relatively high ignition probabilities at low release rates.
88
2.6.7
Offshore risers
The results of the riser scenario simulations show that these typically give maximum ignition probabilities of
0.001 to 0.003. There is little difference between the open and openj configuration results. The results do not
seem significantly sensitive to the riser location, since all but the very small releases would give momentum jets
long enough to reach platform areas.
2.6.8
The results of the platform engulfing simulations show that the ignition probability can reach a maximum of
approximately 0.1. Oil releases with some rain-out give slightly higher ignition probabilities compared with
pure gas releases, since the liquid spills add to the ignition probability, but the differences are not significant.
It should be noted that the Phase 1 report [1] quotes Per Holand, giving a recommended ignition
probability of 0.1 for blowouts for use in QRA based on recent historical data and trends.
2.6.9
Results discussion
A cursory inspection of the results suggests that these do not seem out of step with current QRA data and
assumptions, and the results do appear to vary appropriately with factors such as dimensions, congestion and
environmental setting.
As expected, in most cases the results lie within the Cox, Lees and Ang ignition correlations, and the
offshore scenarios are in reasonable alignment with the 0.02 (2 %) average historical ignition rate for 'QRA'
releases derived from the OIR 12 data analysis.
In addition, as expected, very large gas and LPG releases onshore in industrial areas can reach ignition
probabilities higher than 0.3 (the maximum in Cox, Lees and Ang), with large vapour clouds extending over
areas with relatively high ignition source densities.
2.6.9.1 Sensitivity analysis
Selected sensitivity analysis was carried out for a range of onshore and offshore release scenarios to test the
ignition probability results against a number of factors considered important. The results show that the 'base
case' model results are reasonably robust and do provide a suitable basis for assessing a range of typical
conditions.
Lower windspeed
A series of runs were undertaken to compare results with the base case windspeed of 5 m/s and a lower speed of
2 m/s. The results showed that the lower windspeed does give higher ignition probabilities for most scenarios
for release rates in the 'mid range'. These could increase by approximately a factor of two. However, the
windspeed had little effect on the minimum and maximum ignition probabilities for the scenarios: in essence,
the lower windspeed shifted the mid part of the ignition probability-release rate curve slightly to the left. The
windspeed had no noticeable effect on the scenarios modelled as 'openj' configurations, since these use a
momentum jet model for dispersion.
Higher windspeed
A series of runs were undertaken to compare results with the base case windspeed of 5 m/s and a higher speed
of 8 m/s. The results showed that the higher windspeed does give lower ignition probabilities for most
scenarios for release rates in the 'mid range'. These could decrease by approximately a factor of one and a half
(i.e. pro-rata to the windspeeds). As for the lower windspeed tests, there was little or no effect on the maximum
values reached.
Given that windspeeds at or below 2 m/s are relatively uncommon both onshore and offshore (present
less than 10 % of the time) and that windspeeds at or above 8 m/s (present approximately 25 % and 40 % of the
time for onshore and offshore respectively) give slightly lower ignition probabilities than the 'base case', then a
'base case' 5 m/s wind does appear to provide a suitable overall value that probably errs on the cautious side for
offshore releases.
It should be noted that the look-up correlations are based on an averaged windspeed in order to
represent an averaged ignition probability for use across a wide range of situations in a typical QRA. The
sensitivity results suggest that the value of 5 m/s adopted is reasonable on this basis.
89
In some QRAs, it may be necessary to undertake detailed analysis under specific weather conditions of
interest, including factors such as the wind direction, windspeed, ambient temperatures and weather stability.
These factors can affect the dispersion of any vapour, and hence the probability of that vapour reaching
potential ignition sources. If it is considered that the use of 'averaged' ignition probabilities is not sufficiently
discriminating in these situations, then the full ignition model should be used. This includes a feature to allow
the results of any detailed dispersion modelling to be fed directly into the model (as flammable vapour volumes/
areas in the different Areas).
Offshore flare vs. cold vent for Area 4
The modelling has assumed that attended offshore installations will have a flare system, which can provide an
ignition source for very large releases. This will be conservative for those installations having only a cold vent
system.
Sensitivity analysis was undertaken for a range of representative offshore module gas releases to assess
the difference between a flare vs. cold vent ignition source characteristic for Area 4 (i.e. type 16 Offsite flare
vs. 15 Offsite vent respectively). The results show that for release rates up to 10 kg/s there is little
difference: these do not extend significantly into Area 4 (off the platform edge). Above 10 kg/s, the ignition
probabilities for the releases on an Area 4 'flare' installation do become higher than those for an Area 4 'cold
vent' installation. At these higher release rates, the differences are generally less than a factor of 2, with typical
maximum ignition probabilities of 0.2 for the 'cold vent' situation compared to typical maximum values of 0.035
for the 'flare' situation.
The results show that the 'flare' basis used for developing the look-up correlations is, as expected,
conservative for installations only having a cold vent. However, a flare only adds approximately 0.015 (1.5 %)
to the overall ignition probability for very large releases. This is considered to be within the overall error bands
of ignition probability estimation using the model. It is considered that the look-up correlations developed do
provide a reasonable basis for assessing releases on installations with either a flare or cold vent.
Time spent with hot work
The ignition model includes an allowance for ignition by hot work activity in the plant area. A nominal 15 hr.
of hot work per year was used as the base case. Increasing this to 100 hr. for the sensitivity test (for a selected
number of onshore and offshore representative scenarios) showed that this still only gave a low annual
probability of hot work activity being present. The model results show that there is negligible effect on the
overall ignition probabilities from this level of hot work activity.
Electrical isolation time
The ignition model uses the electrical isolation time to determine the time available for intermittent ignition
sources to ignite any released gas, vapour or liquid in the plant. Different values have been used in the
modelling depending on the type of plant area and the likely detection and emergency shutdown times. A
sensitivity test was conducted by increasing the base case electrical isolation times by a factor of three.
This had a negligible effect on many release scenarios, where the ignition probability may be more
dependent on the chance of ignition in adjacent areas or offsite. An effect was noticeable for some scenarios,
but this was a minor effect in the mid range release rates, and did not affect the maximum ignition probabilities.
Overall, the result can be considered relatively insensitive to the electrical isolation time.
Liquid bund size
The ignition probability of liquid spills can be very dependent on the limiting pool size, generally related to the
bund size. This was confirmed by some preliminary sensitivity analysis and, as a result, the base case modelling
work has addressed different bund/ maximum spill sizes directly in the scenarios.
Material inventory
The inventory of material is used in the model, amongst other things, to limit the extent of any liquid or vapour
dispersion. The inventory can therefore have a significant effect on the ignition probability in some
circumstances. It should be noted that the inventory considered in the model includes the 'locked in' inventory
once the plant has been isolated and any material released up to this point, i.e. the total mass of material
available to be released.
In general, plant releases are not particularly sensitive to the available inventory, since the amount of
material needed to form a flammable cloud in the plant can be relatively small, and this inventory is often
90
provided by the quantity of material released before the detection and shutdown systems activate to isolate the
inventory.
The inventory is more relevant to large leaks from storage facilities, which could spread a long way
offsite. However, the inventories in storage tanks are normally significant (tens of tonnes or more); this is
sufficient to support long hazard ranges.
It was recognised at the onset of the study that the inventory may be a significant factor for storage
area releases, and this has been addressed directly in the modelling, with a range of inventories being assessed.
The model results have shown that the typical ranges of inventories involved are sufficient to support
long hazard ranges, and hence high ignition probabilities. However, the nature of vapour dispersion means that
the results are not that sensitive to the inventory, and the look-up correlations proposed are reasonably
representative for the typical ranges of inventories used. The effect of inventory is further discussed in 2.7.2.6.
2.6.9.2 Gas build-up JIP workbook and flammability limits
The current ignition model uses the Gas Build-up JIP Workbook method [8] to assess the flammable volume
within the Area 1 plant area. This has been based on large-scale tests using a gas with flammability limits of
0.05 to 0.15 vol/vol. The results of the modelling of gas (essentially methane) and LPG for this Phase 2 project
have shown that the standard workbook method can give results which suggest that LPG may form a lower
flammable volume that a methane gas release for a given release scenario. This is due to the effect of the
vapour density term in the workbook equation and the different flammability limits.
In practice, it would be expected that an LPG release would result in a larger flammable cloud volume
than methane since this has relatively wider flammability limits, and a reduced LFL. Further investigation of
the Gas Build-up JIP Workbook method shows that a correction factor is suggested to account for different
LFLs. This correction has been included in Version D1 of the UKOOA Ignition model. However, this
correction does not take into account the effect of the wider flammability range for higher molecular weight
hydrocarbons, e.g. LPG vs. methane. It may therefore, still underestimate slightly the flammable volumes for
some hydrocarbons. This is unlikely to be significant in overall risk terms due to the conservatisms in the Gas
Build-up JIP Workbook as applied to the ignition model: the ignition model uses the more conservative 'upper
estimate' workbook equation rather than the 'typical' equation.
The modelling of dispersion in Areas 2 to 4 in the ignition model uses standard gas dispersion
methods, which do take account of the vapour density and flammable limits. In addition, the 'open' and 'openj'
configuration options for Area 1 use momentum jet correlations, not the Gas Build-up JIP Workbook method.
The momentum jet correlations take full account of the gas molecular weight and LFL. As a result, the majority
of the results generated for the various scenarios are not significantly affected. In addition, the look-up
correlations have been established taking into account both the gas and LPG ignition probability results. It is
therefore considered that the look-up correlations provide a suitable basis for assessing both methane and LPGtype releases even in the light of the shortcomings in the standard Gas Build-up JIP Workbook method.
2.7
LOOK-UP CORRELATIONS
2.7.1
The results of the scenario ignition modelling have been used to derive a number of representative best-fit
curve/ correlations, which can be used in QRA to estimate the probability of ignition for a wide range of
common scenarios.
The correlations consist of up to three gradients (grad_a, grad_b, grad_c), each of the generic form:
Log10 ( y ) = m Log10 ( x ) + c , rearranged as y = 10
[ m Log10 ( x ) + c ]
where y is the ignition probability, x is the mass release rate (kg/s), m is the 'gradient' of the correlation, and c is
the y-axis 'offset' of the correlation.
Gradient a and offset a characterise the correlation between points A and B, gradient b and offset b
between points B and C, and gradient c and offset c between points C and D. If an upper point is not specified,
the gradient and offset values are used for all values of x (the mass flow rate) above the lower point value.
Points A, B, C and D are specified as x values (i.e. mass release rates).
Maximum and minimum values are also assigned.
The correlation values are shown in Table 2.9.
91
The look-up correlations were derived by selecting suitable 'best fit' lines to match or envelope the
model results. Charts showing these 'best fit' lines and the model results they have been selected to represent
are presented in Figure 2.25 through to Figure 2.40.
Charts showing the actual look-up correlations (best fit curves) for all the scenarios are presented in
Figure 2.41 through to Figure 2.45. These have been grouped and presented to enable comparison between
related look-up correlations.
An additional correlation is provided for '30 Tank Liquid diesel, fuel oil' based on the 100 m x
100 m liquid tank correlation, but with a reduced ignition probability to reflect the lower ignition potential for
liquids stored at near ambient conditions below their flash point (see Table 2.10).
The look-up correlations also include equivalents of the commonly used Cox, Lees and Ang ignition
probability correlations [2] for gas and liquid, with maximum values of 0.3 and 0.08 respectively.
A look-up function has also been developed in Microsoft Excel, which allows the look-up
correlations to be called using a simple cell function formula linked to a visual basic module in the Look-up
correlations workbook (Version D1). This workbook includes the full look-up function code in visual basic.
The look-up correlations take the form of a number of log-log gradients chosen to match or envelope
the model results. In some of the larger LPG release cases for pipelines and tanks in open areas, the 'open' and
'openj' configuration modelling gives noticeably different ignition probabilities. The 'open' configuration
modelling results are more representative of an LPG release that causes rain-out and pool formation followed by
rapid evaporation, resulting in relatively passive wind driven dispersion, whereas the 'openj' case would be more
representative of a vapour momentum jet release. In practice, actual releases are likely to exhibit behaviour
somewhere between these extremes: so for these cases, the look-up correlation has taken a mid line between the
'open' and 'openj' results.
2.7.2
In general, the selection of the most suitable correlation for use in a given QRA situation should be
straightforward. However, some guidance is offered below on the basis of the scenarios, some of the key
considerations involved, and selecting the best correlation for some specific situations.
2.7.2.1 Onshore pipelines
The onshore pipeline correlations have conservatively assumed that the release is at or near a valve station or
other location where there may be some equipment or activities within the pipeline corridor that could present a
chance of ignition similar to a process area. The scenario also assumes there are no automatic leak detection or
electrical or process isolation systems along the pipeline route, so the release would continue for some time,
allowing any liquid pool size to build up. This is reflected in the relatively high ignition probabilities for small
pipeline liquid releases compared to plant releases. However, since the ignition probabilities concerned are still
very low, and well below the Cox, Lees and Ang 'liquid' ignition probability values, further evaluation to refine
the correlation has not been considered worthwhile.
2.7.2.2 Tank farms
The best-fit correlation for a gas or LPG storage area within a plant ('14 Tank Gas LPG Storage Plant') is
intended to represent a tank farm within a plant, where plant areas and equipment largely surround the tanks.
This can result in high ignition probabilities due to the relatively high plant area ignition source densities
compared to a tank farm. The correlation is not representative of leaks from tanks within plant areas: the
relevant 'plant' correlations should be used for these situations.
Tank farms located adjacent to plant can generally be treated as being in an industrial area, and it is
recommended that the correlation for storage in an industrial environment be used ('15 Tank Gas LPG Storage
Industrial').
Tank farms in an isolated rural setting should use '16 Tank Gas LPG Storage Rural'.
The 100 m x 100 m bund correlation is intended to be representative of the majority of tank farms and
storage areas. The 300 m x 300 m bund correlation is intended for use in very large storage facilities with
multi-tank bunds, where there could be ignition sources associated with several tanks in the bund.
For a storage area with very large (100 m diameter or more) tanks, then the bund areas should be
relatively free of sources of ignition. For such storage areas with only a few (one to five) very large tanks
sharing a bund, then the 100 m x 100 m ignition correlation should be used.
92
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
0.0024
0.001
0
0
0.001
0.001
0.001
0.001
0.001
0.001
0.04
0.025
0.15
0.15
0.028
0.1
0.3
0.08
0.001
0.001
0.001
0.001
0.001
0.001
0.001
0.001
0.001
0.001
0.001
0.001
0.001
0.001
0.001
0.001
0.001
0.001
0.001
0.001
0.001
min_p
0.0175
0.01
0.025
0.04
0.05
0.07
0.007
1
1
0.6
0.1
0.013
0.65
0.13
0.05
0.7
0.12
0.015
1
1
0.5
Max_p
0.1
0.5
0.5
0.1
0.1
0.1
0.1
0.1
0.1
0.1
0.1
0.1
0.1
0.1
0.1
0.1
0.1
0.1
0.1
0.1
0.1
0.1
0.1
0.1
0.1
0.1
0.1
0.1
0.1
0.1
pointa
1.134699
-1.82391
0.525219
-2.43339
0.072551
-2.88606
0.544175
-2.41443
0.468588
-2.49002
0.652869
-2.30574
For comparison
0.641939
-1.80676
0.392472
-1.88185
Special (derived) scenarios
0.010724
-2.98928
grad_a
offset_a
Onshore scenarios
0.558795
-2.18593
0.605288
-2.13944
0.739652
-2.21896
0.129819
-2.82879
0.356547
-2.60206
0.338819
-2.61979
0.338819
-2.61979
0.356547
-2.60206
0.356547
-2.60206
0.338819
-2.61979
0.356547
-2.60206
0.075721
-2.90309
0.075721
-2.90309
0.023065
-2.93554
0.023065
-2.93554
0.023065
-2.93554
Offshore scenarios
0.400548
-2.55806
0.400548
-2.55806
0.037789
-2.92082
0.768182
-2.19042
0.845058
-2.11355
93
Note: 'N' as a gradient or offset means 'not applicable', 0 as a point also means 'not applicable'.
Type
No.
100
100
1
30
1
0.3
100
100
100
100
1
3
5
70
0.3
1000
10
1
1
1
1
1
1
1
1
1
1
1
1
pointb
0.067994
N
N
0.216588
N
1.213764
1.231261
N
N
N
N
0.880788
0.390377
0.285097
N
0.127125
N
0.80103
1.56813
0.809894
0.809894
1
0.840621
0.809894
1.211604
0.395757
0.395757
1.458907
1.145784
1.123063
grad_b
-2.98928
N
N
-1.82391
N
-2.88606
-2.05517
N
N
N
N
-2.92082
-2.01017
-1.72215
N
-2.38946
N
-3.5
-2.60206
-2.61979
-2.61979
-2.60206
-2.60206
-2.61979
-2.60206
-2.90309
-2.90309
-2.93554
-2.93554
-2.93554
offset_b
0
0
50
0
50
10
0
0
0
0
30
10
30
0
70
0
1000
3
100
100
100
100
100
70
7
7
100
100
10
pointc
0.554906
N
N
N
N
N
N
N
N
N
N
N
N
N
N
N
N
N
0.734824
N
N
N
N
N
0.31737
0.88091
0.88091
N
0.647338
0.40624
grad_c
-3.40076
N
N
N
N
N
N
N
N
N
N
N
N
N
N
N
N
N
-2.20447
N
N
N
N
N
-0.95211
-3.31309
-3.31309
N
-1.93865
-2.21872
offset_c
500
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
1000
0
0
0
0
0
1000
500
500
0
700
1000
pointd
94
Pign Overall = 1 [(1 Pign Gas / LPG ) (1 Pign Liquid )] combined method
If the ignition probability contribution of the 20 kg/s of gas was 0.05 and if the ignition probability contribution
of the 80 kg/s of liquid was 0.02, then the overall ignition probability would be =1-((1-0.05)x(1-0.02)) = 0.069.
95
The ignition model calculates the liquid ignition contribution based on a simple estimate of the pool depth and
assuming a vapour layer above the liquid surface: both the estimates of the pool spread and the extent of the
vapour layer are considered conservative. Also, the area covered by the vapour and liquid is likely to overlap
considerably, so the two ignition probability contributions are not strictly additive. On this basis, it is
considered that the higher of the two ignition probabilities ('vapour' and 'liquid') could be adopted for QRA as a
reasonable estimate of the overall ignition probability.
5
6
Application
Onshore scenarios
Releases of flammable liquids that do not have any significant flash
fraction (10 % or less) if released from onshore cross-country
pipelines running through industrial or urban areas
Releases of flammable liquids that do not have any significant flash
fraction (10 % or less) if released from onshore cross-country
pipelines running through rural areas
Releases of flammable gases, vapour or liquids significantly above
their normal (Normal Atmospheric Pressure (NAP)) boiling point
from onshore cross-country pipelines running through industrial or
urban areas
Releases of flammable gases, vapour or liquids significantly above
their normal (NAP) boiling point from onshore cross-country pipelines
running through rural areas
Releases of flammable gases, vapour or liquids significantly above
their normal (NAP) boiling point from small onshore plants (plant area
up to 1200 m2, site area up to 35,000 m2)
Releases of flammable liquids that do not have any significant flash
fraction (10 % or less) if released from small onshore plants (plant
area up to 1200 m2, site area up to 35,000 m2) and which are not
bunded or otherwise contained
Releases of flammable liquids that do not have any significant flash
fraction (10 % or less) if released from small onshore plants (plant
area up to 1200 m2, site area up to 35,000 m2) and where the liquid
releases from the plant area are suitably bunded or otherwise
contained
Releases of flammable gases, vapour or liquids significantly above
their normal (NAP) boiling point from large onshore outdoor plants
(plant area above 1200 m2, site area above 35,000 m2)
Releases of flammable liquids that do not have any significant flash
fraction (10 % or less) if released from large onshore outdoor plants
(plant area above 1200 m2, site area above 35,000 m2) and which are
not bunded or otherwise contained
96
97
98
99
For the majority of scenarios, the 'highest value' approach and 'combined' approach will probably give
very similar results. Either could be used.
2.8
2.8.1
One of the assessments required, when calculating the risks from releases of flammable material, is the
probability distribution for the various possible ignited outcomes. These could include one or more of a variety
of hazards such as a flash fire, pool fire, jet fire, fireball or explosion. The relative chances of these occurring
given ignition can have a significant effect on the risks estimated by the analysis. Explosions, for example, can
generate overpressures capable of causing major escalation and personnel injury over a wide area, whereas a
flash fire may only have localised effects.
Various factors can affect the outcome from an ignited leak, including:
The timing of ignition can also be important, especially when evaluating the time for personnel to escape an
event, or when considering the ignition of large drifting flammable vapour clouds, which have the potential to
affect large areas.
The uncertain and complex nature of the various factors involved in ignition means that risk
assessment practitioners have developed simplified rules and assumptions for ignition modelling.
Many quantified risk assessments (QRAs) include factors, within event trees for example, to:
Often these approaches make the simplistic assumption that 'early ignition' would result in a jet fire (momentum
vapour or 2-phase leak) or pool fire (liquid leak) or fireball (large catastrophic leak of vapour), whereas delayed
ignition of vapour or aerosol is typically taken to result in a flash fire or explosion, depending on the congestion
in the area concerned.
Historical accident data provide an indication of the relative incidences of ignited event types, such as
various types of fires and explosions; however, these data are too generic to be of relevance to specific
individual scenarios needing assessment in a QRA.
Two factors may help improve the estimates used to determine the relative distribution of ignited events:
How best to interpret existing historical ignited event outcome data for use in QRA?
Would the timing of ignition provide a basis to estimate the nature of the ignited outcome?
These two factors are discussed below, drawing on various reviews and analyses of ignition data from both
offshore and onshore sources.
2.8.2
Ignition timing
The Phase 1 report [1] included an analysis of historical data on ignition probability and ignition timing from
numerous sources as part of an overall study on ignition modelling. It concluded that data on the timing of
ignition and the resulting type of flammable event are sparse.
For offshore blowouts, the review indicated that approximately 40 % ignite immediately, with almost
50 % occurring over an hour after the blowout started [3].
Onshore MAH data derived from an analysis of LPG incidents in an accident database indicate that
30 % of ignitions occurred within 30 s of the leak and 70 % within 2 min. [4]. These data are presented in Table
2.11 for plant, transport and pipeline situations.
100
1
0.22
0.53
0.24
0.10
0.89
Note 1 The OIR 12 data are dominated by small leak and ignition events that ignited 'immediately': most of
these would not be relevant to MAH QRA. The 'Plant' distribution in the table may be more representative of
an ignition timing distribution for major leaks on an offshore installation.
By contrast, approximately 90 % of the process release ignitions reported in the OIR12 1992-2003 data were
classified as igniting immediately [5] (see Table 2.12). This may reflect the nature of many of the ignitions in
the OIR12 data related to internal turbine or flare/vent events or hot work activity rather than large process
leaks spreading to meet ignition sources. It is expected that typical QRA MAH would have a higher probability
of delayed ignition, more in line with the 'plant' data in Table 2.11.
Table 2.12 OIR 12 data ignition timings
No. Ignitions
46
10
10
66
No. Immediate
41
10
8
59
No. Delayed
5
0
2
7
% Immediate
89.1
100.0
80.0
89.4
% Delayed
10.9
0.0
20.0
10.6
Table 2.11 presents an overview of ignition timing data. Assuming early ignition relates to ignition within one
minute or so of the leak commencing, then the typical QRA practice to assume a 50:50 or 30:70 distribution
between early and delayed ignitions appears to be reasonable given the data available. Based on the data in
Table 2.11, a 30:70 (early: delayed) distribution may be representative of early ignition within 30 s, whereas a
50:50 distribution may be representative of early ignition within 60 s.
The OIR12 data indicate that early ignition may be relatively more likely than the typical 50:50 or
30:70 split, however assuming these higher probabilities of delayed ignition is considered conservative: this
would give more severe consequences, e.g. an initial explosion or flash fire followed by a jet or pool fire,
although it may also imply more time for people to escape the area.
The UKOOA Ignition model also provides an estimate of the ignition probability distribution by time
and location, based on the ignition probability contribution and timing of ignition at the following points:
The timing of ignition is sometimes used as a means to predict the nature of the ignited event. Early ignition is
taken to indicate a jet fire or pool fire depending on the material concerned (gas or liquid). Delayed ignition is
taken to indicate that the ignition would initially result in a flash fire or explosion.
101
Some analyses assume a historically based flash fire to explosion ratio to determine the relative
frequency of these events. However, this does not take account of the specific layout of the area in terms of its
confinement and congestion, which affects the chance of an explosion developing.
A review of offshore fire and explosion data from WOAD indicates a 21 % incidence of an explosion
given ignition [4].
Cox, Lees and Ang [2] have suggested a mass release rate based correlation for an explosion given
ignition ranging from 0.04 for small gas releases, 0.12 for medium gas releases, to 0.3 for massive gas releases.
The process release ignition data reported in the OIR12 1992-2003 data are presented in Table 2.13 [5].
Table 2.13 OIR 12 data ignition outcome distribution by media
Process ignitions
Gas
Condensate
Oil
Total
% Jet fires
29
8
18
24
% Pool fires
0
50
27
13
% Flash fires
54
33
55
51
% Explosions
17
8
0
13
These data suggest an averaged historical incidence of explosions of approximately 20 % for gas releases. The
OIR 12 data also indicate a high proportion of flash fires. Interrogation of the raw data suggests most of these
were associated with small leaks or spills during hot work activity: these may not be relevant to typical QRA
major leak scenarios.
The analysis of the ignited process leak events was repeated, classified by leak severity, to remove the
'minor' leaks, which would not be considered as relevant to MAH QRA. The results are shown in Table 2.14.
Since there have been no 'major' event ignitions, the 'significant' data set offers the most relevant data.
Table 2.14 OIR 12 data ignition outcome distribution by leak severity
Severity analysis
Minor
Significant
Major
Total
No. Ignitions
43
23
0
66
No. Immediate
40
19
0
59
No. Delayed
3
4
0
7
% Immediate
93.0
82.6
N/A
89.4
% Delayed
7.0
17.4
N/A
10.6
Of the 23 ignited significant process leaks, one was a condensate incident in a flare, four were oil leaks, of
which one was a flare incident and one was a power turbine incident; the remainder (18 leaks) were of gas, of
which seven were power turbine incidents, and five were vent/ flare incidents. This leaves only two oil leaks
and six gas leaks of any direct relevance to QRA type process leak ignitions.
The two oil leak ignitions were classed as immediate, and neither resulted in an explosion (outcomes
were a flash fire and pool fire respectively). Of the six gas leak ignitions, only two were classed as delayed, but
two of the incidents (classed as immediate ignition) involved an explosion. There were also two flash fire and
two jet fire outcomes, with the two delayed ignitions leading to the jet fires. This gives two out of eight events
classified as 'delayed' ignition, i.e. 25 %. Of these events, only two were associated with normal operation as
opposed to maintenance or restart conditions. It is therefore concluded that the data do not provide a sound
basis on which to draw any firm conclusions on the ignition timing distribution or end event outcome
distribution for MAH QRA leaks.
There may be some subjectivity in what is classed as 'immediate ignition' for the OIR12 database, but it
is likely that this implies ignition very soon after the release started, probably within a few seconds.
This difference between the relative probability distribution of early vs. delayed ignition and the
distribution between the ignited outcomes in the OIR 12 data is further highlighted when Table 2.12 is
compared with Table 2.13.
2.8.4
The findings of the reviews of data on ignition timing show that approximately 30 % to 50 % of ignitions occur
within a minute or so of the leak commencing. The typical QRA assumption is that these early ignitions lead to
pool or jet type fires, and that explosions are associated with delayed ignition.
102
However, investigation of the OIR 12 data shows that of the nine explosion events, eight were classified as
immediate ignition and only one was associated with delayed ignition. Most flash fires, pool fires and jet fires
were classified as immediate ignition. However, two of the 'significant' jet fire events were associated with
delayed ignition.
This suggests that:
The timing of the ignition, i.e. immediate/ early or delayed is not a reliable indicator of the type of ignited
event that may subsequently ensue. Events classified as 'immediate' or 'early' ignitions could lead to
explosions, and events classified as 'delayed' ignitions may result in no reported explosion or flash fire.
There are too little data to draw any firm conclusions on the relative probabilistic distribution of fire/
explosion outcomes given ignition.
The high probability of immediate ignition from the OIR 12 data may be due to the way the events were
detected and the nature of the many hot work, turbine internal ignitions and vent/flare events in the data set.
The OIR 12 ignition timing distribution may not be representative of the ignition event outcome distribution for
MAH QRA type leaks and ignitions. Nevertheless, the data indicate that simple assumptions, e.g. explosions of
gas clouds are associated with events classified as 'delayed ignition', may not be valid.
2.8.5
The review of ignition timing and ignited event outcome data has shown that there are limited data available,
insufficient to support detailed distributions for scenario specific ignition timing or the relative probabilities of
fires and explosions. The reported timing of ignition is also shown to be an unreliable indicator of the type of
fire or explosion event that could result from the ignition.
From a practical risk assessment perspective, it is therefore suggested that the hazardous outcomes
from an ignition event are predominantly based on the physical phenomena associated with the release:
1.
2.
3.
4.
5.
6.
7.
Assumptions 1-4 are simply related to the physical leak type and are independent of the timing of ignition.
Even if these events were to include an explosion, this would be followed by e.g. a jet fire or pool fire until the
leak source could be isolated.
2.8.5.1 Time dependent ignition probabilities
If a time dependent distribution is required, for example to assess the ability of persons to escape the event or
take mitigation measures, then the time distributions in Table 2.11 could be adopted. From available data, it is
suggested that the 'Plant' distribution is the most appropriate for assessing MAH QRA leaks from both onshore
and offshore plant (see Table 2.15).
Table 2.15 Plant ignition timings distribution
Time, t (s)
Plant
1000
0.94
> 1000
1.0
Alternatively, the UKOOA Ignition model could be used to estimate the ignition distribution by time and
location.
103
Day
Night
Average
These ignition source density values assume all ignition sources are continuous and have a certainty of ignition,
i.e. a probability of ignition or ignition potential of 1.0 given contact with a flammable vapour. Under these
conditions, the probability of non-ignition, Q(t), at some time, t, is given by:
Ln Q(t) = - A(t)
or
where:
A(t) is the area extent of the cloud within flammable limits (hectares) at time, t
, is the ignition source density for the land use type within the cloud (sources per hectare)
The probability of ignition at any point in time, P(t), is given by:
P(t) = 1- Q(t)
or
i.e. 1.0 minus the overall probability of non-ignition, or 1.0 minus the product () of the probabilities of nonignition for all Areas within the cloud.
This approach to assessing the probability of ignition with time as the cloud spreads can be a useful means of
determining an approximate ignition timing/ consequence distribution to use in a QRA. In practice, it may be
useful to consider the distribution as a number of steps, for example, representing ignition on plant, at the site
boundary, at the nearest offsite residential areas and at the cloud's fullest extent. The 'Plant' ignition time
distribution in Table 2.15 could be used in conjunction with the typical windspeed and plant dimensions to
estimate the probability of early ignition on plant. The rural/ urban/ industrial ignition source densities could
then be used to assess the delayed ignition distribution off site.
Alternatively, the UKOOA Ignition model could be used to estimate the ignition distribution by time
and location, but this only gives ignition probability contributions at a few points: these may not be suitable for
clouds drifting some distance offsite.
2.8.5.3 Probability of an explosion
The probability of an explosion (Item 6 above) is more complex to assess, since this depends on many scenario
specific factors. Confinement and congestion in the area are needed to generate the overpressures, but factors
such as the ignition type and location and the concentration of the vapour can also affect the chance of ignition
and the burning regime in the cloud.
A simple distribution such as that suggested by Cox, Lees and Ang [2], or a simple fraction, e.g. 20 %
from the data review (see 2.8.3), could be adopted for relatively generic screening risk assessments. However,
104
the detailed risk assessment of a particular facility may wish to review/ modify these simple factors depending
on the nature of the leaks and plant configurations being assessed.
For scoping risk assessments, one of the following two approaches should be adopted:
Conservative 'worst-case' assume that all ignitions of gas/ vapour/aerosol in confined or congested areas
result in explosions.
Indicative generic case assume that 20 % of all ignitions of gas/ vapour/ aerosol in confined or congested
areas result in explosions or use a simple mass release rate based correlation (e.g. that in Cox, Lees and
Ang) to assess the probability of an explosion given ignition. Some qualitative discussion/ justification
may be needed to show why the range of situations being addressed can be considered 'typical' or 'average'
in terms of their probability of an explosion.
The conservative worst-case approach could also be used in more complex, detailed and high-risk/ high
uncertainty situation assessments, but this may lead to gross overestimates of the risks.
In these situations, some form of probabilistic explosion modelling (e.g. using exceedance curves)
may be needed to determine the relative probability distribution for different explosion overpressures/ impulses
based on the modelling of different release source terms, ignition strengths, timings and locations, etc.
One approach to assessing the probability of an explosion in these cases would be to adjust the
historical 'average' ignition probability based on the nature of the scenario. As a simple rule of thumb, areas
having relatively low congestion and confinement or few potential ignition sources could be expected to have
explosion probabilities (given ignition) of less than the 'average' of 20 % (0.2), say in the range 0.1 to 0.2.
Areas with high congestion and confinement or many potential ignition sources could be expected to have
explosion probabilities (given ignition) of greater than the 'average' of 20 % (0.2). A value of 30 % (0.3) (the
maximum in the Cox, Lees and Ang correlation) would probably be appropriate for a relatively congested or
confined plant or module. Given that an explosion requires a narrower range of flammable limits, ignition
energy and local configuration compared to that for a general ignition, the probability of an explosion given
ignition would be expected, in practice, to be significantly less than 100 % (1.0), and would probably be
unlikely to exceed 50 % (0.5) even in very favourable (for an explosion) release and ignition configurations.
An alternative approach would be to take the data from Table 2.15, which indicate that approximately
25 % of ignitions would occur within the first few seconds of the leak. These can be considered too 'early' to
allow a gas cloud to develop and spread sufficient to give a major flash fire or explosion. This leaves
approximately 75 % of events with flash fire or explosion potential. The relative probabilities and severities of
the various flash fires and explosion outcomes could then be assessed using probabilistic explosion modelling
techniques.
In practice, it is likely that these two approaches would give similar results for the overall probability
of an explosion.
105
Phase 1 report: Ignition probability review and model development, which forms part of IP Research
Report Ignition probability review, model development and look-up correlations, Energy Institute
(January 2006), ISBN 978 0 85293 454 8 .
Classification of hazardous locations, Cox, Lees and Ang, Institution of Chemical Engineers (May 1991),
ISBN 0 85295 258 9.
Offshore blowouts, causes and trends, Per Holand, Norwegian Institute of Technology, Trondheim,
Norway (1996).
A review of ignition probabilities for use in offshore installation quantified risk assessments, AEA
Technology, HSE OSD Research Project 3209 (May 1995) (unpublished internal HSE report).
HSE OIR12 on-line database, http://www.hse.gov.uk/hcr3/ (accessed January 2004).
A model for the ignition probability of flammable gases (Phase 2), WS Atkins, HSE Contract Research
Report 203/1998, HSE Books (1998), ISBN 0 7176 1657 6.
Development of a method for the determination of on-site ignition probabilities, WS Atkins, HSE Research
Report 226, HSE Books (2004), ISBN 0 7176 2843 4.
Gas build up from high pressure natural gas releases in naturally ventilated offshore modules, JIP
workbook on gas accumulation in a confined and congested area, BP Amoco, CERC, BG Technology
(May 2000) (referred to as the Gas Build-up JIP Workbook).
A guide to quantitative risk assessment for offshore installations, J Spouge, CMPT, Publication 99/100
(1999), ISBN 1 870553 365.
106
107
1
OilpipeJ_indust
Oil pipeline leaks as jet release in
industrial area
OilpipeJ_rural
Oil pipeline leaks as jet release in
rural area
Ignition Probability
0.1
OilpipeJ10%flash_indust
Oil pipeline leaks with 10 % flashing
as jet release in industrial area
OilpipeJ10%flash_rural
Oil pipeline leaks with 10 % flashing
as jet release in rural area
0.01
Oilpipe_indust
Oil pipeline leaks in industrial area using
Gas Build-up JIP dispersion model
0.001
Oilpipe_rural
Oil pipeline leaks in rural area using
Gas Build-up JIP dispersion model
0.0001
0.1
Oilpipe10%flash_indust
Oil pipeline leaks in industrial area using
Gas Build-up JIP dispersion model with
10 % flashing of oil
1
10
100
1000
Flowrate kg/s
Oilpipe10%flash_rural
Oil pipeline leaks in rural area using Gas
Build-up JIP dispersion model with 10 %
flashing of oil
0.1
Ignition Probability
LPGpipeJ_indust_test1_Lo Pressure
LPGpipe_indust
LPG pipeline leaks in industrial area using
Gas Build-up JIP dispersion model
LPGpipe_rural
0.001
LPGpipe_indust_test1_Lo Pressure
LPG pipeline leaks using Gas Build-up
JIP dispersion model at low pressure as
sensitivity test
0.0001
0.1
10
100
1000
Flowrate kg/s
108
1
GaspipeJ_indust
Gas pipeline leaks as jet release in
industrial area
GaspipeJ_rural
Gas pipeline leaks as jet release in
rural area
Ignition Probability
0.1
GaspipeJ_indust_test1_igntime
Sensitivity test to electrical isolation time
10 min. vs. zero
GaspipeJ_indust_test2_length
Sensitivity test to plant length used:
100 vs. 10 m
0.01
Gaspipe_indust
Gas pipeline leaks in industrial area using
Gas Build-up JIP dispersion model
Gaspipe_rural
Gas pipeline leaks in rural area using Gas
Build-up JIP dispersion model
0.001
Gaspipe_indust_test1_igntime
Sensitivity test to electrical isolation time
-10 min. vs. zero
0.0001
0.1
10
Gaspipe_indust_test2_length
1000 Sensitivity test to plant length used:
100 vs. 10 m
100
Flowrate kg/s
Small_open_gas_indust
Gas leaks in small open/unconfined
plant in industrial area
Small_open_LPG_indust
LPG leaks in small open/unconfined
plant in industrial area
0.1
Ignition Probability
Small_open_Oil10%flash_indust
Oil leaks with 10 % flashing in small
open/unconfined plant in industrial area
Small_open_liq_indust
Liquid leaks in small open/unconfined
plant in industrial area
0.01
Small_open_gas_rural_bund
Gas leaks in small open/unconfined
plant in rural area (bund not relevant)
Small_open_LPG_rural_bund
LPG leaks in small open/unconfined
plant in rural area with bund to limit
pool spread
0.001
Small_open_Oil10%flash_rural_bund
Oil leaks with 10% flashing with bund
to limit pool spread
0.0001
0.1
10
100
Small_open_liq_rural_bund
Oil leaks with no flashing with bund
1000 to limit pool spread
Flowrate kg/s
Figure 2.6 Model results for small onshore plant Open plant area
109
Small_tight_gas_indust
Gas leaks in small confined plant in
industrial area
Small_tight_LPG_indust
LPG leaks in small confined plant in
industrial area
0.1
Ignition Probability
Small_tight_Oil10%flash_indust
Oil leaks with 10% flashing in small
confined plant in industrial area
Small_tight_liq_indust
Liquid leaks in small confined plant
in industrial area
0.01
Small_tight_gas_rural
Gas leaks in small confined plant in
rural area
Small_tight_LPG_rural
LPG leaks in small confined plant in
rural area
0.001
Small_tight_Oil10%flash_rural
Oil leaks with 10% flashing,
rural location
0.0001
0.1
10
100
Small_tight_liq_rural
1000 Oil leaks with no flashing, rural
location
Flowrate kg/s
Figure 2.7 Model results for small onshore plant Confined/ congested plant area
1
Large_open_gas_indust
Gas leaks in large open/unconfined
plant in industrial area
Large_open_LPG_indust
LPG leaks in large open/unconfined
plant in industrial area
Ignition Probability
0.1
Large_open_Oil10%flash_indust
Oil leaks with 10 % flashing in large
open/unconfined plant in industrial
area
Large_open_liq_indust
Liquid leaks in large open/unconfined
plant in industrial area
0.01
Large_open_gas_rural_bund
Gas leaks in large open/unconfined
plant in rural area (bund not relevant)
Large_open_LPG_rural_bund
LPG leaks in large open/unconfined
plant in rural area with bund to limit
pool spread
0.001
Large_open_Oil10%flash_rural_bund
Oil leaks with 10% flashing with bund
to limit pool spread
0.0001
0.1
10
100
1000
Large_open_liq_rural_bund
Oil leaks with no flashing with bund
to limit pool spread
Flowrate kg/s
Figure 2.8 Model results for large onshore plant Open plant area
110
Large_tight_gas_indust
Gas leaks in large confined plant
in industrial area
Large_tight_LPG_indust
LPG leaks in large confined plant
in industrial area
0.1
Ignition Probability
Large_tight_Oil10%flash_indust
Oil leaks with 10 % flashing in large
confined plant in industrial area
Large_tight_liq_indust
Liquid leaks in large confined
plant in industrial area
0.01
Large_tight_gas_rural
Gas leaks in large confined plant
in rural area
Large_tight_LPG_rural
LPG leaks in large confined plant
in rural area
0.001
Large_tight_Oil10%flash_rural
Oil leaks with 10% flashing
0.0001
0.1
10
100
1000
Large_tight_liq_rural
Oil leaks with no flashing
Flowrate kg/s
Figure 2.9 Model results for large onshore plant Confined/ congested plant area
1
LPG_20t_Plant_indust
0.1
LPG_20t_Plant_rural
Ignition Probability
0.01
LPG_2000t_Plant_rural
0.001
LPG_2000t_Store_indust
0.0001
0.1
10
100
Flowrate kg/s
1000
LPG_2000t_Store_rural
Figure 2.10 Model results for onshore tank farm LPG Passive dispersion
111
Oil_5000t_10%flash_Plant_indust
Oil_5000t_10%flash_Plant_rural
0.1
Oil_50000t_10%flash_Plant_rural
Ignition Probability
0.01
0.001
Oil_5000t_Store_Indust
0.0001
0.1
10
100
Flowrate kg/s
Figure 2.11 Model results for onshore tank farm Oil Passive dispersion
1
LPG_20t_Plant_indust
0.1
Ignition Probability
0.01
LPG_2000t_Plant_rural
0.001
LPG_2000t_Store_indust
0.0001
0.1
10
100
Flowrate kg/s
1000
LPG_2000t_Store_rural
Figure 2.12 Model results for onshore tank farm LPG Release treated as momentum jet
112
Oil_5000t_10%flash_Plant_indust
Oil (10%flash) releases from 5000t storage
tank in area within plant in industrial area
Oil_50000t_10%flash_Plant_indust
As above but 50,000t tank
Oil_5000t_10%flash_Plant_rural
As above but plant in rural area
0.1
Ignition Probability
Oil_50000t_10%flash_Plant_rural
As above but 50,000t tank
Oil_5000t_Plant_indust
As above but no flashing and plant in an
industrial area
Oil_5000t_Plant_rural
As above but rural area
0.01
Oil_5000t_10%flash_Store_Indust
As above but storage area situated away
from plant or storage only
Oil_5000t_10%flash_Store_rural
As above but storage in a rural area
0.001
Oil_5000t_Store_Indust
As above but storage in an industrial area
Oil_5000t_Store_rural
As above but no flashing and storage in a
rural area
Oil_5000t_Store_Indust_nobund
As above but no bund-so pool allowed to
spread further, and industrial location
0.0001
0.1
10
100
1000
Flowrate kg/s
Oil_5000t_Store_rural_nobund
As above but rural location for storage
Figure 2.13 Model results for onshore tank farm Oil Release treated as momentum jet
1
Confined_Process_oil_100flash
100% flashing oil release in confined
process module
Confined_Process_oil_15partflash
15% flashing oil release in confined
process module
Ignition Probability
0.1
Confined_Process_oil_noflash
No flash oil release in confined process
module
0.01
Process_oil_100flash
100% flashing oil release in more typical
process module
0.001
Process_oil_15partflash
15% flashing oil release in more typical
process module
0.0001
0.1
10
100
1000
Process_oil_noflash
No flash oil release in more typical
process module
Flowrate kg/s
Figure 2.14 Model results for offshore process module Oil releases
113
Wellhead_oil_100flash
100% flashing oil release in
wellhead module
Ignition Probability
0.1
0.01
Wellhead_oil_15partflash
15% flashing oil release in
wellhead module
0.001
0.0001
0.1
10
100
1000
Wellhead_oil_noflash
No flash oil release in
wellhead module
Flowrate kg/s
NUI_gas
Gas release on NUI process deck
Ignition Probability
0.1
NUI_Liq_allflash
100% Flashing Oil/ liquid release
on NUI process deck
0.01
NUI_liq15flash
15 % Flashing Oil/ liquid release
on NUI process deck
0.001
NUI_liq_noflash
0.0001
0.1
10
100
Flowrate kg/s
114
1000
FPSO_gas
Gas release on FPSO process deck
FPSO_Liq_allflash
Ignition Probability
0.1
FPSO_liq15flash
15 % Flashing Oil/ liquid release
on FPSO process deck
0.01
FPSO_liq_noflash
No Flash Oil/ liquid release on FPSO
process deck
0.001
FPSO_gas_1000
Gas release on FPSO process deck
Gas inventory limited to 1000 kg
FPSO_gas_500
0.0001
0.1
10
100
1000
Flowrate kg/s
Ignition Probability
0.1
FPSO_gas_wall
Gas release on FPSO process deck behind
(downwind of) transverse
wall across deck
FPSO_Liq_allflash_wall
As above but 100 % flashing oil
or liquid release
0.01
FPSO_liq15flash_wall
As above, but 15 % flashing oil
or liquid release
0.001
0.0001
0.1
10
100
1000
Flowrate kg/s
Figure 2.18 Model results for offshore FPSO process module behind transverse wall
115
Confined_Process_gas
Gas release in confined process module
Confined_Process_LPG
LPG release in confined process module
Ignition Probability
0.1
Process_gas
Gas release in more typical process
module
Process_LPG
LPG release in a more typical process
module
Large_Process_gas
Gas release in large process module
0.01
Process_gas_mech15
Gas release in process module with
mechanical ventilation of 15 acph
Process_LPG_mech15
LPG release in process module with
mechanical ventilation of 15 acph
0.001
Process_gas_mech30
Gas release in process module with
mechanical ventilation of 30 acph
0.0001
0.1
10
100
1000
Flowrate kg/s
Process_LPG_mech30
LPG release in process module with
mechanical ventilation of 30 acph
Process_gas_opendeck
Gas release from process on open deck
Figure 2.19 Model results for offshore process module Gaseous releases
Wellhead_gas
Gas release in wellhead module
Wellhead_LPG
LPG release in wellhead module
Ignition Probability
0.1
Confined_Proc_compresn_gas
Gas release in confined compression
module
Confined_Proc_compresn_LPG
LPG release in confined compression
module
0.01
Proc_compresn_gas
Gas release in more typical compression
module
0.001
Proc_compresn_LPG
LPG releasein more typical compression
module
Confined_Proc_compresn_oil15
_flash
15 % flashing oil release in confined
compression module
0.0001
0.1
10
100
Flowrate kg/s
1000
Confined_Proc_compresn_oil
No flash oil release in confined
compression module
Figure 2.20 Model results for offshore wellhead and compression modules Gaseous releases
116
riser_nearedge_openj_gas
Gas jet release from riser in air gap near
edge of platform
riser_nearedge_openj_LPG
LPG jet release from riser in air gap near
edge of platform
0.1
riser_nearedge_openj_oil15_flash
Ignition Probability
riser_nearedge_openj_oil_noflash
As above but no flashing
0.01
riser_nearedge_open_gas
Gas release from riser treated as jet near
riser, but as passive dispersion thereafter
riser_nearedge_open_LPG
LPG release from riser treated as jet near
riser, but as passive dispersion thereafter
0.001
riser_nearedge_open_oil15_flash
As above but 15% flashing oil release
riser_nearedge_open_oil_noflash
As above but no flashing
0.0001
0.1
10
100
1000
Flowrate kg/s
Figure 2.21 Model results for offshore riser Release near to edge of installation
riser_openj_gas
riser_openj_LPG
LPG jet release from riser in air gap away
from edge of platform
0.1
riser_openj_oil15_flash
Ignition Probability
riser_openj_oil_noflash
riser_open_gas
Gas release from riser treated as jet near
riser, but as passive dispersion thereafter
riser_open_LPG
LPG release from riser treated as jet near
riser, but as passive dispersion thereafter
0.001
riser_open_oil15_flash
As above but 15% flashing oil release
riser_open_oil_noflash
0.0001
0.1
10
100
Flowrate kg/s
1000
riser_openj_gas_big_airgap
Sensitivity test using larger air gap
Figure 2.22 Model results for offshore riser Release away from edge of installation
117
0.1
Ignition Probability
offshore_engulf_gas
0.01
offshore_engulf_Oil_flash_
15
As above, but 15 % flashing oil
or liquid release
0.001
0.0001
0.1
10
100
1000
Flowrate kg/s
Ignition Probability
0.1
0.01
Cox_lees_Ang_gas
Cox_lees_Ang_oil
0.001
Cox_lees_Ang_gas_unlimited
Cox_lees_Ang_oil_unlimited
0.0001
0.1
10
100
Flowrate kg/s
Figure 2.24 Cox, Lees and Ang ignition probabilities for comparison
118
1000
NUI_gas
NUI_Liq_allflash
0.1
NUI_liq15flash
NUI_liq_noflash
0.01
Cox, Lees, Ang - Gas
0.001
0.1
10
100
1000
Process_gas
Process_LPG
Wellhead_gas
Wellhead_LPG
0.1
Proc_compresn_gas
Proc_compresn_LPG
Large_Process_gas
Process_gas_opendeck
0.01
0.001
0.1
10
100
1000
Figure 2.26 Look-up vs. model results for offshore process module Gas
119
Confined_Process_gas
Confined_Process_LPG
Confined_Proc_compresn_gas
Confined_Proc_compresn_LPG
0.1
Confined_Proc_compresn_oil15_flash
Process_gas_mech15
Process_LPG_mech15
Process_gas_mech30
0.01
Process_LPG_mech30
0.001
0.1
10
100
1000
Figure 2.27 Look-up vs. model results for offshore process module Gas Congested or mechanically
ventilated module
1
riser_openj_gas
riser_openj_LPG
riser_openj_oil_noflash
riser_open_gas
riser_open_LPG
riser_open_oil_noflash
0.1
riser_nearedge_openj_gas
riser_nearedge_openj_LPG
riser_nearedge_openj_oil_noflas
h
riser_nearedge_open_gas
riser_nearedge_open_LPG
riser_nearedge_open_oil_noflas
h
riser_openj_gas_big_airgap
0.01
BF Offshore Riser
Cox, Lees, Ang - Gas
Cox, Lees, Ang - Liquid
0.001
0.1
10
100
1000
120
FPSO_gas
FPSO_Liq_allflash
FPSO_liq15flash
0.1
FPSO_liq_noflash
FPSO_gas_1000
FPSO_gas_500
0.01
0.001
0.1
10
100
1000
Figure 2.29 Look-up vs. model results for offshore FPSO no wall
1
FPSO_gas_wall
FPSO_Liq_allflash_wall
0.1
FPSO_liq15flash_wall
0.01
0.001
0.1
10
100
1000
Figure 2.30 Look-up vs. model results for offshore FPSO with transverse wall
121
offshore_engulf_gas
offshore_engulf_Oil_flash_50
0.1
offshore_engulf_Oil_flash_15
BF Offshore Engulf_blowout_riser
0.01
Cox, Lees, Ang - Gas
0.001
0.1
10
100
1000
Figure 2.31 Look-up vs. model results for offshore release engulfing platform
Confined_Process_oil_noflash
Process_oil_15partflash
Process_oil_noflash
Wellhead_oil_15partflash
0.1
Wellhead_oil_noflash
Proc_compresn_oil_15_flash
Proc_compresn_oil
Wellhead_oil_15_flash
Wellhead_oil
0.01
Process_oil_noflash_mech15
BF Offshore Process Liquid
Cox, Lees, Ang - Gas
Cox, Lees, Ang - Liquid
0.001
0.1
10
100
1000
Figure 2.32 Look-up vs. model results for offshore process module Liquid
122
GaspipeJ_indust
GaspipeJ_rural
LPGpipeJ_indust
LPGpipeJ_rural
0.1
Gaspipe_indust
Gaspipe_rural
LPGpipe_indust
LPGpipe_rural
BF Pipe Gas LPG Industrial
0.01
0.001
0.1
10
100
1000
Figure 2.33 Look-up vs. model results for onshore pipelines Gas/ LPG
OilpipeJ_indust
OilpipeJ_rural
0.1
Oilpipe_indust
Oilpipe_rural
0.001
0.1
10
100
1000
Figure 2.34 Look-up vs. model results for onshore pipelines Liquid
123
Small_tight_gas_indust
Small_tight_LPG_indust
Small_tight_gas_rural
0.1
Small_tight_LPG_rural
Small_open_gas_indust
Small_open_LPG_indust
Small_open_gas_rural_bund
0.01
Small_open_LPG_rural_bund
0.001
0.1
10
100
1000
Figure 2.35 Look-up vs. model results for onshore small plant Gas/ LPG
Small_tight_Oil10%flash_indust
Small_tight_liq_indust
Small_tight_Oil10%flash_rural
Small_tight_liq_rural
0.1
Small_open_Oil10%flash_indust
Small_open_liq_indust
Small_open_Oil10%flash_rural_bund
Small_open_liq_rural_bund
0.01
BF Small Plant Liquid
0.001
0.1
10
100
1000
Figure 2.36 Look-up vs. model results for onshore small plant Liquid
124
1
Large_tight_gas_indust
Large_tight_LPG_indust
Large_tight_gas_rural
Large_tight_LPG_rural
0.1
Large_open_gas_indust
Large_open_LPG_indust
Large_open_gas_rural_bund
Large_open_LPG_rural_bund
0.01
0.001
0.1
10
100
1000
Figure 2.37 Look-up vs. model results for onshore large plant Gas/ LPG
1
Large_tight_Oil10%flash_indust
Large_tight_liq_indust
Large_tight_Oil10%flash_rural
Large_tight_liq_rural
0.1
Large_open_Oil10%flash_indust
Large_open_liq_indust
Large_open_Oil10%flash_rural_bun
d
Large_open_liq_rural_bund
0.01
0.001
0.1
10
100
1000
Figure 2.38 Look-up vs. model results for onshore large plant Liquid
125
LPG_20t_Plant_indust
LPG_20t_Plant_rural
LPG_20t_Store_indust
0.1
LPG_20t_Store_rural
openj-LPG_20t_Plant_indust
openj-LPG_20t_Plant_rural
openj-LPG_20t_Store_indust
openj-LPG_20t_Store_rural
BF Tank Gas LPG Storage Plant
0.01
BF Tank Gas LPG Storage Industrial
BF Tank Gas LPG Storage Rural
Cox, Lees, Ang - Gas
Cox, Lees, Ang - Liquid
0.001
0.1
10
100
1000
Figure 2.39 Look-up vs. model results for onshore storage Gas/ LPG
Oil_5000t_Plant_indust
Oil_5000t_Store_Indust
0.1
Oil_5000t_Store_rural
Oil_5000t_Store_Indust_nobund
Oil_5000t_Store_rural_nobund
0.01
0.001
0.1
10
100
1000
Figure 2.40 Look-up vs. model results for onshore storage Liquid
126
Ignition Probability
0.01
0.001
0.1
10
100
1000
Figure 2.41 Look-up correlations (best fit) for onshore liquid releases
Ignition Probability
0.01
0.001
0.1
10
100
1000
Figure 2.42 Look-up correlations (best fit) for onshore gas/LPG releases
127
0.1
Ignition Probability
BF Offshore Riser
10
100
1000
Figure 2.43 Look-up correlations (best fit) for offshore liquid releases
1
BF Offshore Process Gas
Opendeck NUI
BF Offshore Process Gas
Typical
BF Offshore Process Gas
Large Module
Ignition Probability
0.1
0.001
0.1
10
100
1000
Figure 2.44 Look-up correlations (best fit) for offshore gas releases
128
BF Offshore Riser
0.1
BF Offshore Engulf_blowout_riser
0.01
0.001
0.1
10
100
1000
Figure 2.45 Look-up correlations (best fit) for offshore engulfing release and riser release
129
130