PLOFAM Report

Working together
for a safer world
Process leak for offshore

installations frequency assessment
model - PLOFAM
Main report
Report for:
Statoil Petroleum AS
Report no: 105586/R1 Rev: Final B

Date: 18 March 2016
Summary
Process leak for offshore installations frequency assessment model - PLOFAM
Main report
Security classification of this report: Open distribution
Report no: Revision: Report date:

105586/R1 Final B 18 March 2016
Prepared by: Reviewed by: Approved by:

Ingar Fossan/Are Opstad Sb Erik Odgaard Kristin Myhre
Senior Principal Consultant/ Senior Principal Consultant Department Manager
Senior Consultant
Contributing contractor: Contributing contractor: Contributing contractor:
Jon Andreas Hestad/Geir Drage

Jo Wiklund/Jens Morten Nilsen
Berentsen
Principal Advisor/Senior Analyst
Senior Safety Engineer/Safety
Engineer
Entity name and address: Client name and address:

Lloyd's Register Consulting - Energy AS Statoil Petroleum AS
Fjsangerveien 50B Postboks 8500
NO-5059 BERGEN NO-4035 STAVANGER
Norway Norway
Our contact: Client contact:

Are Opstad Sb Eli Glittum
T: +47 476 68 619 T: +47 915 43 176
E: are.sabo@lr.org E: elgl@statoil.com
Lloyds Register Group Limited, its subsidiaries and affiliates and their respective officers, employees or agents are, individually and
collectively, referred to in this clause as Lloyds Register. Lloyds Register assumes no responsibility and shall not be liable to any
person for any loss, damage or expense caused by reliance on the information or advice in this document or howsoever provided,
unless that person has signed a contract with the relevant Lloyds Register entity for the provision of this information or advice and
in that case any responsibility or liability is exclusively on the terms and conditions set out in that contract.
Except as permitted under current legislation no part of this work may be photocopied, stored in a retrieval system, published,
performed in public, adapted, broadcast, transmitted, recorded or reproduced in any form or by any means, without the prior
permission of the copyright owner. Enquiries should be addressed to Lloyds Register, 71 Fenchurch Street, London, EC3M 4BS.
Lloyds Register 2016.
Report no: 105586/R1 Rev: Final B Page i

Date: 18 March 2016 Lloyds Register 2016
Report no: 105586/R1 Rev: Final B Page ii
Document history
Revision Date Description/changes Changes made by
Draft 09.11.2015 First version Ingar Fossan
Final 08.12.2015 The report is extensively restructured and updated, Ingar Fossan and
including all technical notes Are Opstad Sb
Final A 09.12.2015 Revised executive summary and conclusion. Model Ingar Fossan and Are
parameters included also in this main report. No Opstad Sb
changes to the data basis or model. TN-3, TN-5
and TN-6 are re-issued as Final A
Final B 18.03.2016 Revised based on comments from Statoil. TN-1, TN- Ingar Fossan and Are
2 and TN-4 are re-issued as Final A, while TN-3, TN- Opstad Sb
5 and TN-6 are re-issued as Final B.
Executive summary
Hydrocarbon process leaks are a major contributor to offshore risk. The last decade the industry has used
a model denoted Offshore QRA - Standardised Hydrocarbon Leak Frequencies (SHLFM) to estimate leak
frequencies for these incidents. This model originates from the JIP project Standardised Hydrocarbon
Leak Frequencies, which was first reported in final version in 2005. Based on experience from use of the
model, Statoil has appreciated the need for a thorough revision of the methodology, and initiated a
project where the purpose has been to create an updated leak frequency model that can be accepted as
an industry standard for the Norwegian Continental Shelf by consultancy companies and operators.
To achieve this, Statoil contracted Lloyds Register Consulting (LRC), DNV GL, Safetec and Lilleaker
Consulting AS to work together. In addition to the four consultancy companies, the operators
ConocoPhillips and Lundin were invited to the project. LRC has been the lead contractor while the others
have contributed as advisors through workshops, document review and discussions in meetings. The
project has been run during the period March December 2015. Personnel participating in workshops are
listed in Table 1.1. Also other persons have been involved in video conferences and discussions.
This report with technical notes documents the resulting leak frequency model, denoted PLOFAM (Process
leak for offshore installations frequency assessment model) that will be the preferred model by all above
mentioned project participants. It is expected that this model will be used for all QRAs for Statoil,
ConocoPhillips and Lundin.
The leak frequency model covers process leaks and topside leaks from the well system occurring during
normal production. The leak scenarios may have a leak point associated with well, process or utility
systems. The leak frequency for process leaks estimated by the model accounts for leaks occurring both in
the process system and utility system fed from the process system. The model does however not give
separate leak frequencies for process releases through utility systems and through process system. Three
main leak scenarios are defined for the leak frequency model. That is Process leak, Producing well leak
and Gas lift well leak. Furthermore, the model distinguishes on leak scenarios where the total released
amount of hydrocarbons is 10 kg, and >10 kg. These leaks are classified as Marginal leaks and
Significant leaks, respectively. Only the Significant leak scenario is relevant for detailed modelling of
consequences and dimensioning accidental loads in a formal QRA. The Marginal leak scenario is only
relevant with regard to immediate exposure of personnel in the close vicinity to scene of the leak to
accidental loads.
Report no: 105586/R1 Rev: Final B Page iii

The model itself consists of a set of mathematical equations that generate the frequency hole size
distribution per standard equipment type per equipment dimension. Thus, the model is highly equipment
size dependent, and a significant effort has been made to build a model where both the total leak
frequency and the frequency for ruptures is equipment size dependent, and unique for every standard
equipment type. The model includes the following new equipment types not included SHLFM; compact
flanges, steel piping, flexible piping and a model for leaks from hoses used in temporary operations.
The model is thoroughly validated towards available data of leaks that has occurred at installations both
on the United Kingdom Continental Shelf (UKCS) and the Norwegian Continental Shelf (NCS). A main
overall conclusion obtained from running the parameterization and validation process is that the
underlying hole size frequency distribution for equipment at installations located on the NCS is similar to
the distribution for equipment located on UK installations. The differences may be explained by
uncertainty related to the datasets (both the leaks and the population data), limitations of the mathe-
matical formulations and uncertainty associated with the parameterization process. The model validation
shows that PLOFAM is able to:
(1) Reproduce the total number of leaks at NCS in the period 01.01.2001 31.12.2014
(2) Reproduce the observed distribution of leaks with respect to initial leak rate for installations on the
NCS. The model predicts the individual contribution from Significant and Marginal leaks as well as the
individual frequency distribution for both leak scenarios
(3) Reproduce the observed contribution to leaks originating from the different equipment types. The
model does also reproduce the observed frequency distribution of leaks with respect to initial leak
rate for the most dominating equipment types at NCS (i.e. valves, flanges, instruments and steel
pipes).
The strategy has been to build a model that gives a best estimate for future leak frequencies, i.e. to create
an unbiased model without built in conservatism. It is observed a significant decreasing trend in historical
leak frequency with time for installations on the NCS in the period after year 2000. The period 2001-2014
is used as basis for the model, but the historical frequency for all installations on NCS the last 5 year
period is 40% less than the average for the period 2001-2014. Hence, presuming that the number of
leaks at the NCS in the future will follow the observed frequency for the last 5 year period, PLOFAM is
regarded to give robust results for future average leak frequencies for installations on the NCS. The period
2001-2014 is used as basis for the model to account for uncertainty in the data material and shifts in
underlying casual factors (e.g. emerging unknown degradation mechanisms due to age or changing
operational conditions) affecting the future trend in leaks occurring on installations on the NCS.
It should be mentioned that the historical leak frequency per installation at the NCS can vary significantly
from the NCS average, as a result of stochastic effects, and also if the conditions at a particular installation
deviates from the normal conditions at installations on NCS.
The main outstanding shortcoming of the model is the capability to reflect the relative contribution from
liquid and gas leaks. PLOFAM is not able to predict the individual contribution from gas and liquid leaks
accurately. This is most likely caused by uncertainties related to the actual phase of the leaking medium,
which influences both how historical incidents are logged, and how initial leak rates are modelled in the
validation process. The validation model shows that PLOFAM will underpredict the number of small gas
leaks and overpredict the number of small liquid leaks for the period 2001-2014, which has been put
down as basis for the model development. The total sum of gas leaks and liquid leaks in the period is
predicted accurately. For the period after 2007, PLOFAM overpredict the frequency for both gas and liquid
leaks over the entire spectre of initial leak rates. The shortcoming is suggested to be addressed in future
projects, for instance as part of the future project updating PLOFAM.
Report no: 105586/R1 Rev: Final B Page iv

Table 1.1 Personnel participating in workshops. Also other persons have been involved in video
conferences and discussions
Name Company Role
Eli Glittum Statoil Statoil project manager
Unni Nord Samdal Statoil Technical point of contact
Espen Fyhn Nilsen Statoil Technical point of contact
Marie Saltkjel ConocoPhillips Participant
Espen Skilhagen Lundin Participant
Robert Schumacher Lundin Participant
Are Opstad Sb Lloyds's Register Consulting Project manager/participant
Ingar Fossan Lloyds's Register Consulting Technical responsible
Erik Odgaard Lloyds's Register Consulting Quality assurer
Jan Pappas Lloyds's Register Consulting Participant
Jens Garstad DNV GL Participant
Andreas Falck DNV GL Participant
Jo Wiklund Lilleaker Consulting AS Participant
Jens Morten Nilsen Lilleaker Consulting AS Participant
Jon Andreas Hestad Safetec Participant
Geir Drage Berentsen Safetec Participant
Morten Skjong Safetec Participant
Ole Magnus Nyheim Safetec Participant
Glossary/abbreviations
Abbreviations and expressions used in the main report and all technical notes are given in TN-1.
Abbreviations relevant for the main report are repeated in Table 1.2. An important expression, frequently
used in the model, is the Complementary cumulative hole size frequency distribution. This expression
denotes frequency distributions F(hole size > d), where d is a specific hole size. This expression is
throughout the report denoted , and for simplicity it is referred to as the hole size frequency distribution.
Note that the complementary cumulative hole size probability distribution for an equipment type
multiplied by the total leak frequency for that equipment type, gives the complementary cumulative hole
size frequency distribution.
Table 1.2- Abbreviations used in main report and technical notes
Abbreviation Description
ASCV Annulus safety check valve
ASV Annulus safety valve
DHSV Downhole safety valve
ESD Emergency shut down
Report no: 105586/R1 Rev: Final B Page v

HCRD Hydrocarbon release database
HSE Health and safety executive
LRC Lloyds Register Consulting
Modelling of ignition sources on offshore oil and gas
MISOF
Facilities
NCS Norwegian continental shelf
Process Leak for Offshore installations Frequency
PLOFAM
Assessment Model
P&ID Piping and instrumentation diagram/drawing
Ptil Petroleumstilsynet (Petroleum safety authority)
PWV Production wing valve
QRA Quantitative risk analysis
Risikoniv i norsk petroleumsvirksomhet (Risk level in
RNNP
Norwegian petroleum industry)
SHLFM Standardised hydrocarbon leak frequencies model
TN Technical note
UKCS United kingdom continental shelf
Report no: 105586/R1 Rev: Final B Page vi

Table of contents Page
1 Introduction .....................................................................................................................................1
1.1 Report structure ....................................................................................................................1
1.2 Objective ...............................................................................................................................1
2 Leak scenarios covered by the model ...............................................................................................1
3 Model summary and application of the model .................................................................................5
3.1 Model summary ....................................................................................................................5
3.2 Application of the model .......................................................................................................7
4 Equipment types included in the model ...........................................................................................7
5 Data basis ........................................................................................................................................8
5.1 NCS data ...............................................................................................................................8
5.2 UKCS data .............................................................................................................................9
6 Model parameterisation and validation ............................................................................................9
7 PLOFAM parameters ......................................................................................................................12
8 Comparison PLOFAM with SHLFM .................................................................................................13
9 MISOF benchmarking model and PLOFAM.....................................................................................15
10 Robustness of PLOFAM ..................................................................................................................16
11 Concluding remarks.......................................................................................................................17
12 References .....................................................................................................................................19
Report no: 105586/R1 Rev: Final B Page vii

1 Introduction
This report describes the leak frequency model, denoted PLOFAM (Process leak for offshore
installations frequency assessment model), and used for estimation of topside process leak
frequencies for use in Quantitative Risk Analysis of fire and explosion at installations located on
the Norwegian Continental Shelf.
The model is in general fully documented in the technical notes (TN) listed in Chapter 1.1. This
main report presents the most important aspects of the model, without presenting all details, but
with a sufficient level of detail to gain an overview of the model. For further details it is referred
to the TNs throughout the report.
1.1 Report structure

The report consists of the following technical notes:
TN-1 Expressions and abbreviations
TN-2 NCS data
TN-3 UKCS data
TN-4 Leak scenarios
TN-5 Leak frequency model
TN-6 Parametrization of model parameters and validation.
1.2 Objective
The objective of the leak frequency model is to predict the future leak frequency for topside
process leaks at installations located on the Norwegian Continental Shelf (NCS).
2 Leak scenarios covered by the model

The leak frequency model covers process leaks and topside leaks from the well system occurring
during normal production. A detailed description of system boundaries and scenarios covered by
the model is given in TN-4.
The leak scenarios covered by the model may have a leak point associated with well, process or
utility systems, and are described in Table 2.1. Other leak scenarios, such as leaks from utility
systems fed from utility systems (for example diesel from diesel tanks and MEG from MEG-
system) are not included.
Note that the leak frequency for process leaks estimated by the model does also account for
leaks occurring in the utility system, but being fed from the process system. This is done by
including process leaks fed through utility systems, but not equipment counts from utility systems
as basis for the model validation. This implies that utility equipment should not be counted as
basis for estimation of process leak frequencies. Furthermore the model does not give separate
leak frequencies for process releases through utility systems and through process system. This
means that a QRA based on PLOFAM will not reflect the potential location of the leak sources in
utility systems. Furthermore, the leak frequency contribution from utility systems will scale with
the number of equipment counts for process system. This contribution will in practice vary
somewhat with the system at hand, but this cannot be quantified based on PLOFAM. A detailed
risk assessment of leaks in utility systems, if found required, should hence be covered by special
evaluations. Figure 2.2 gives an illustration of leak scenarios normally considered in a QRA. The
figure shows which scenarios that are covered by the model and which that are not.
Report no: 105586/R1 Rev: Final B Page 1

Incidents occurring during well interventions/operations, such as wire line and coiled tubing, are
defined as blowouts or well releases, and are covered by Ref. /1/ that is based on the SINTEF
Offshore Blowout Database. These incidents are not covered by the model.
Table 2.1 - Leak scenarios covered by the model. They occur in well system, process system or utility
system (process leaks fed through utility systems). Scenarios that are not listed in this table are not
covered by the model
Leak point in well system Leak point in process Leak point in utility system
system
1. Producing well/Injection 4. Leak point in 5. Leak point in flare system (low
well: Topside well release process system pressure or high pressure flare
where the inventory bet- between PWV and system)
ween DHSV and PWV is topside riser ESDV/- 6. Excessive releases through flare
released during normal storage ESDV. The tips and atmospheric vents that
production. fuel system is exceed the design specification
2. Gas lift well: Topside well regarded as part of and pose a fire and explosion
release where the inven- the process system. hazard to equipment, structures
tory between the ASV or personnel. Such leaks are de-
and the barrier towards noted vent leaks.
the process system is 7. Leak point in utility systems that
released. In cases where is fed by hydrocarbons stemm-
no ASV is present, the ing from process system.
entire inventory in the gas Systems covered by the model
lift annulus to the ASCV are:
may be released. Assu-
a. Open drain system
ming that the check valve
ASCV is functioning, b. Closed drain system
otherwise there is no c. Chemical injection systems.
barrier towards the reser-
voir.
3. Release of hydrocarbon
fluid from annuli that are
not used for gas lift.
Three main leak scenarios for modelling in QRAs are defined in PLOFAM. That is Process leak,
Producing well leak and Gas lift well leak.
For all leak scenarios, 0.1 kg/s is recommended as the general leak rate threshold for estimation
of leak duration (both in terms of calculation of fluid dispersion and fire duration) in a QRA, for
3
all leak scenarios in open areas and leaks in enclosures having a net volume more than 1,000 m
and with ventilation rate of 12 ach or higher (see TN-4). The lower leak rate threshold is put as
basis for the lower boundary with regard to aggregated released amount of hydrocarbons (10
kg). The model distinguishes on leak scenarios where the total released amount of hydrocarbons
is 10 kg, and >10 kg. These leaks are classified as Marginal leaks and Significant leaks,
respectively.
In a QRA, the risk in terms of fire- and explosion load exposure to vulnerable equipment and
structures such as safety systems, pressurized equipment, load carrying structures and main
safety functions, associated with Marginal leaks can be neglected. However, the risk to personnel
associated with Marginal leaks should not be neglected.
The three main leak scenarios for modelling in QRAs are summarized in Table 2.2, and in Figure
2.1. The table shows how the three main leak scenarios for modelling in QRAs relate to the leak

scenarios in Table 2.1. The figure shows how the leak scenarios in PLOFAM relate to the leak
scenarios in SHLFM, Ref. /2/.
Table 2.2 - Leak scenarios suggested for QRAs at NCS
Modelled leak scenario Leak scenarios included

Significant Scenario 4-7 in Table 2.1, released quantity >10 kg
Process leak
Marginal Scenario 4-7 in Table 2.1, released quantity 10 kg
Production well Significant Scenario 1 and 3 in Table 2.1, released quantity >10 kg
1
leak Marginal Scenario 1 and 3 in Table 2.1, released quantity 10 kg
Gas lift well leak Significant Scenario 2 and 3 in Table 2.1, released quantity >10 kg
Marginal Scenario 2 and 3 in Table 2.1, released quantity 10 kg
Figure 2.1 - Illustration and summary of the leak scenarios to be modelled in a QRA based on
PLOFAM, together with the leak scenarios Full pressure leaks, Limited leaks and Zero pressure leaks
defined in the SHLFM Ref. /2/
1
The frequency for production wells and injection wells are considered to be identical. The leak scenario is
denoted production well only

Figure 2.2 - Illustration of leak scenarios normally considered in a QRA. The figure shows which scenarios that are covered by the model and which that
are not

3 Model summary and application of the model
PLOFAM is based on the assumption that the leak frequency is proportional to the number of
each type of equipment. This assumption is justified in TN-5, where it is also assessed to what
extent the level of operational activity on an installation contributes to leaks. However, for
reasons presented in TN-5, the number of equipment (for each equipment type) is the only
explanatory variable assumed in the model.
This chapter summarizes the mathematical formulation of the model. Furthermore, guidelines for
application of the model and for equipment counting used as input to QRAs are given. The
rational and detailed description of the model is given in TN-5.
3.1 Model summary

The most general formulation of the mathematical equations for the complementary cumulative
hole size frequency distribution, which for simplicity is referred to as the hole size frequency
distribution (see TN-1), is:
(1)
() () , 1 <
(, ) = 0
0 , >
Where 0 is the total leak frequency, is the equipment diameter, is the hole size, and () is
the slope parameter. The equation describes a power law relation that is valid for hole sizes less
than the equipment diameter. The total leak frequency 0 and the full bore hole fraction are
modelled using the following relations.
0 () = ( = 1, ) = () = (0 0 ) (2)
() = ( = , ) = 0 () () = 0 () ( + ) (3)
where the parameters in the equations are described in Table 3.1. A list of PLOFAM parameter
values for , 0 , 0 , , and necessary to estimate leak frequencies for all equipment
types are given in TN-6 Appendix A, and repeated in Chapter 7 below. The slope parameter
(), follows from the assumption of a power law relation and the values for 0 , and :
(4)
log( ) log( + )
0
() = =
() log()
Note that this formulation is the general formulation for all equipment types. For several
equipment types, many of the parameters are set to 0 or 1, resulting in a simpler formulation for
that particular equipment type.
The model is valid for hole sizes 1 mm. The 1 mm hole size diameter threshold is tied to the
lower leak rate threshold of 0.1 kg/s through the validation process. Hence, in cases where leaks
less than 0.1 kg/s may generate significant consequences (e.g. small enclosures), hole diameters
less than 1 mm should be evaluated if that affects the conclusions. The contribution from small
holes is generally expected to be most prominent for liquid leaks.

Neither the data in HCR nor the data gathered from installations at the NCS can be used to
conclude on the behaviour of the distribution for such small holes. It may be that the slope of the
distribution changes when the hole size approaches zero opposed to the slope of the distribution
for bigger holes. However, the parameterized mathematical formulation applied has
demonstrated capability to reproduce the observed distribution of leaks having an initial leak rate
equal to or larger than 0.1 kg/s to the release rate generated in case of rupture of the equipment
(i.e. full bore hole). This is an argument for that the distribution also are able to represent the
underlying behaviour of the distribution of holes having a diameter less than 1 mm. Further
investigations, and preferably access to data of leaks stemming from holes having a diameter less
than 1 mm, is required to investigate further.
Table 3.1 - Summary of all parameters used for each equipment type in the model. The
complementary cumulative hole size frequency distribution, is for simplicity referred to as the hole
size frequency distribution (see TN-1)
Parameter Description
(, ) Hole size frequency distribution (see TN-1) [year-1 component-1].
0 Total leak frequency [year-1 component-1].

0 = ( = 1).
The total full bore hole frequency [year-1 component-1]

= ( = ).
Hole size diameter [mm]
Equipment diameter [mm]
Slope parameter
The average leak frequency (independent of equipment diameter) for the

-1 -1
relevant equipment type [year component ]
0 Parameter in equation for total hole frequency
Parameter in equation for full bore hole frequency
() Adjustment factor of the total leak frequency relative to the weighted

average leak frequency
() Full bore hole fraction

3.2 Application of the model
When applying the model on a specific installation the first step is to define the desired leak rate
intervals defined by the leak rates 1 < 2 < < . Next the following procedure should be
followed for all equipment types and for each relevant equipment diameter D on the installation:
1. Calculate 0 () for the relevant equipment
2. Calculate () for the relevant equipment
4. Calculate the hole size intervals defined by 1 < 2 < < corresponding to the defined
leak rate intervals 1 < 2 < < based on appropriate equations for modelling of leak
rate. TN-6 describes how this is done in the validation process. The models used for
estimating release rates should be carefully chosen based on fluid composition and process
conditions (e.g. pressure and temperature). Since the model is defined for hole sizes >1 mm
only, it is recommended to set to 1 mm if the calculated hole size is < 1mm. Thus leak
frequencies for hole sizes less than 1 mm are not included. This will in general not affect
results in most QRAs as leaks around 1 mm will produce small release rates (< 0.1 kg/s). In
some cases, leaks having a release rate less than 0.1 kg/s ought to be assessed in the QRA to
model the risk picture with adequate precision (e.g. enclosures with poor ventilation). A
special assessment of leaks with an initial leak rate less than 0.1 kg/s has to be performed in
such cases. The HCR data of leaks having hole size 1 mm or less (see TN-3) can be used as
basis for such a study.
5. Calculate leak frequencies for all hole sizes 1 , 2 , for the relevant equipment based on
the equations given in Chapter 3.1
6. Calculate the frequency ( < < +1 ) = ( ) (+1 ) for the hole size intervals and/or
leak rate intervals
7. Multiply the leak frequencies with the number of equipment for the relevant equipment
type. The number of equipment on an installation should be estimated based on equipment
counting on P&IDs or similar. A guideline for counting of equipment is given in TN-5,
Appendix A.
Note that in QRAs based on PLOFAM, no adjustment of calculated leak frequencies should be
performed to take limited operational time into account (see TN-6, Chapter 11.3).
4 Equipment types included in the model

In total 20 different equipment types are covered by the model, including Gas lift well and
Production well, which belong to the well system. The remaining equipment types are the most
common process equipment types at offshore installations. All equipment types covered by the
model are given in Table 4.1.
Table 4.1 - Equipment types included in the model
Equipment type Additional description

Air-cooled heat exchanger
Atmospheric vessel Vessels with atmospheric pressure
Centrifugal compressor
Centrifugal pump
Compact flange

Filter
Flexible pipe
Hose Temporary hoses
Instrument
Pig trap Pig launchers and pig receivers
Plate heat exchanger
Process vessel Pressurized process vessels
Reciprocating compressor
Reciprocating pump
Shell and tube side heat Includes both shell side heat exchangers and tube side heat
exchanger exchangers
Standard flange Includes all flange types, except compact flanges
Steel pipe Process steel pipe
Includes all types of valves. It is recommended to use Valve
(not distinguish on manual, actuated and ESD valve).
Parameters for manual valve, actuated valve and ESD valve
Valve are provided for special studies (see TN-6).
Gas lift well Well head with gas lift
Producing well Well head with or without gas lift
5 Data basis
The model has been developed, parameterised and validated towards data gathered from two
sources of data:
NCS data: 222 incidents recorded at all installations located on the NCS in the period
01.01.2001 31.12.2014
UKCS data: 4561 incidents at installations on the UKCS recorded in HCR database in the
period Q3 1992 - Q1 2015.
The NCS and UKCS databases are described in detail in TN-2 and TN-3, respectively. A short
review is given in Chapter 5.1 and 5.2.
5.1 NCS data

Population data has been collated for 68 installations based on equipment counts extracted from
the QRAs for the installations. 6 out of the 68 installations have not been set in operation by
31.12.2014. Hence, population data is available for 62 installations being in operation in the
whole or part of the period 01.01.2001 31.12.2014.
The data of recorded leaks at the NCS have been established based on the following data sources
1. RNNP dataset collated by Petroleumstilsynet (Ptil) and Safetec
2. Review of accident investigation reports. Accident investigation reports have been available
for the major fraction of the incidents.

Recorded leaks at NCS have all an initial hydrocarbon leak rate of 0.1 kg/s or larger. The total
number of leaks occurring in the period 01.01.2001 31.12.2014 is 222. After review of the
incidents, it has been concluded that 31 of those incidents are not relevant for the leak scenarios
to be modelled, i.e. they are not process leaks or topside well leaks during normal operation (see
Chapter 2 and TN-4). Typical properties of disregarded incidents are as follows:
The leak is a release through a vent or a dump line where the rate is not considered to
exceed the design specification for the vent or dump line
The leak is originating from a piece of equipment not being relevant, such as a pipeline or a
riser
The leak is occurring in the well system during a drilling or workover operation.
Out of the remaining 191 leaks (222 - 31), 181 have occurred at either of the 62 installations in
the NCS population dataset. The remaining 10 leaks have occurred at either of the additional
12 installations in the Full NCS population dataset. Detailed information about all 191 relevant
leaks is given in TN-2 Appendix A.
5.2 UKCS data

Information about offshore releases of hydrocarbons at United Kingdom Continental Shelf
(UKCS), are collected in Hydrocarbon Release Database (HCRD). The database is operated by
Health and Safety Executive (HSE).
In total 4561 events occurring in the period Q3 1992 - Q1 2015 are recorded in HCRD. Not all of
the incidents are relevant for the defined leak scenarios (see TN-4). A thorough analysis has been
necessary to extract the relevant incidents for the model. The resulting databasis contains 2855
recorded incidents from the period Q3 1992 - Q1 2015, and 1597 recorded incidents from the
period Q1 2001 - Q1 2015 that are fed through process systems. By also taking process leaks fed
through utility systems and topside leaks from well systems during normal operation into account,
the total number of leaks put as basis for the model is 3208 (Q3 1992 - Q1 2015).
6 Model parameterisation and validation

The general methodology is illustrated in Figure 6.1. The method consists of three main steps.
The first step in developing the parameters has been to establish initial model parameters solely
based on available data gathered from installations on the UKCS. The only data originating from
NCS that has been utilized to perform the parameterization of the initial parameters is the
relative equipment size distribution taken from the NCS population dataset (see TN-2). The
resulting parameters after performing the initial parameterization have then been validated in
detail towards the data gathered from installations on the NCS in order to understand whether
the model parameters should be adjusted to improve the models ability to predict the occurrence
of leaks on installation located on the NCS.
In order to analyse the difference in underlying frequency distributions for the different equip-
ment types, a second model alternative is developed and validated where it is attempted to
obtain the best possible fit to the NCS data.
The conclusive PLOFAM parameters are based on a methodology that weights the two available
datasets from installations on NCS and UKCS.
Some types of equipment have been subjected to special evaluations where the methodology for
parameterization deviates slightly from the general methodology described above. This is either
due to lack of data in either of the datasets or shortcomings in the available data. The types of
equipment subjected to special evaluations and the reason for the alternative approach is
summarized in Table 6.1.

Step 1: Initial parameterisation
Step 2: Model validation
Establishment of
initial model Step 3: Model development
parameters Validation of
initial model PLOFAM model
based on data Development of
gathered from towards data recommended
installations on gathered from model based on
UKCS (see TN-3) installations on UKCS and NCS
NCS (see TN-2)
data.
Figure 6.1 - Main steps in parameterisation process
Table 6.1 - Equipment types subjected to separate parameterisation process
Equipment Description
type
Hoses No data available in HCRD on leaks occurring under temporary operations

involving use of hoses. The defined model parameters are therefore solely
based on data gathered from installations on the NCS
Steel pipe The quality of the population data in HCRD for steel pipes is judged to be
poor. Hence, the model is parameterized based on a subset of the NCS
population dataset where equipment counts of length steel pipe are
available. However, available data in HCRD have been used to model the
effect of equipment size on the hole size distribution for steel piping
Compact No data is available from UKCS and only limited data available from
flanges installations on NCS. A separate assessment is performed to set the model
parameters
The validation model has demonstrated that PLOFAM is able to:

Generate the total number of leaks at NCS in the period 01.01.2001 31.12.2014 (see
Figure 6.2)
Generate the split of the total number of leaks at NCS in the period 01.01.2001
31.12.2014 on significant and marginal leaks
Generate the number of significant gas leaks at NCS in the period 01.01.2007 31.12.2014
Generate the observed distribution of leaks with respect to initial leak rate for installations on
the NCS. The model predicts the individual contribution from significant and marginal leaks
as well as the frequency distribution for both leak scenarios

Generate the observed contribution to leaks originating from the different equipment types.
The model does also reproduce the observed distribution of leaks with respect to initial leak
rate for the equipment types most of the leaks at NCS are associated with (i.e. valves,
flanges, instruments and steel pipes).
The main outstanding shortcoming of the model is the capability to reflect the relative
contribution from liquid and gas leaks. PLOFAM is not able to predict the individual contribution
from gas and liquid leaks accurately. This is most likely caused by uncertainties related to the
actual phase of the leaking medium, which influences both how historical incidents are logged,
and how initial leak rates are modelled in the validation process. The validation model shows that
PLOFAM will underpredict the number of small gas leaks and overpredict the number of small
liquid leaks for the period 2001-2014, which has been put down as basis for the model
development. The total sum of gas leaks and liquid leaks in the period is predicted accurately. For
the period after 2007, PLOFAM overpredict the frequency for both gas and liquid leaks over the
entire spectre of initial leak rates. The shortcoming is suggested to be addressed in future
Figure 6.2 PLOFAM: Complementary cumulative leak frequency distribution of observed leaks at
NCS and model prediction per equipment type for all installations in the NCS population dataset
(181 leaks observed at 62 installations being in operation in the period 01.01.2001-31.12.2014). The
frequency is obtained by dividing the number of leaks by the number of installation years. The
total number of installation years for the NCS population dataset is 802.

7 PLOFAM parameters
A list of PLOFAM parameter values for , 0 , 0 , , and necessary to estimate leak frequencies for all equipment types are given in
Table 7.1. is given both for Significant and Marginal leaks.
Table 7.1 PLOFAM model parameters. See also TN 6 Appendix A. is given both for Significant and Marginal leaks.
,Significant ,Marginal
Equipment type 0 0
Air-cooled heat exchanger 1.00 - - - 2.70E-02 5.67E-04 -

Atmospheric vessel 1.00 - - - 1.50E-01 2.00E-03 -
Centrifugal compressor 1.00 - - - 1.20E-02 1.50E-03 -
Centrifugal pump 1.00 - - - 3.10E-04 1.90E-03 -
Compact flange 1.00 - - - 1.00E-03 3.30E-06 -
Filter 1.00 - - - 1.00E-02 2.10E-03 -
Flexible pipe 1.00 - - - 7.80E-02 2.50E-04 -
Gas lift well 1.00 - - - 1.70E-02 1.00E-04 1.00E-04
Hose 1.00 - - - 2.00E-01 1.00E-04 2.00E-05
Instrument 0.30 0.40 3.45 -1.25 8.60E-04 1.40E-04 2.00E-05
Pig trap 1.00 - - - 1.20E-02 2.55E-03 -
Plate heat exchanger 1.00 - - - 5.00E-03 6.20E-04 -
Process vessel 1.00 - - - 5.40E-02 3.60E-04 -
Producing well 1.00 - - - 3.80E-02 1.50E-05 9.00E-05
Reciprocating compressor 1.00 - - - 4.20E-02 7.40E-03 -
Reciprocating pump 1.00 - - - 5.60E-02 9.90E-04 -
Shell and tube heat exchanger 1.00 - - - 2.70E-02 1.64E-04 -
Standard flange 0.17 0.40 1.67 -1.50 1.50E-02 3.30E-05 4.50E-06
Steel pipe 4.20 -0.30 17.60 -1.75 2.30E-03 2.50E-05 3.40E-06
Valve 0.19 0.40 2.65 -1.50 6.60E-04 2.74E-04 3.50E-05

8 Comparison PLOFAM with SHLFM
PLOFAM, is compared with the current frequency model commonly used in industry in Norway,
denoted SHLFM (Ref. /2/).
The resulting frequency distributions for the NCS population dataset (62 installations located on
the NCS) obtained by use of PLOFAM and SHLFM are shown in Figure 8.1 and Figure 8.2. The
results display that the difference between the two models is considerable. The difference is
following from:
PLOFAM is based on historical data of leaks occurring at installations on the NCS in the
period 2001-2014. The last version of SHLFM was solely based on data of leaks occurring at
installations on the UKCS in the period 1992-2010. There has been a considerable decrease
in historical leak frequency at installations on the UKCS over this period. It has been shown
that the underlying hole size frequency distribution on the UKCS and NCS after year 2000 is
similar, and most likely the same statement is valid for the total leak frequency as well
Enhanced understanding of the quality of the data in the HCR database, which has provided
basis for implementing considerably less conservatism to account for uncertainty related to
the data basis. This has been facilitated by that more data related to the leaks in HCRD has
been made available by HSE. Two items have been particularly important in this regard.
Firstly, the actual hole size for incidents where the hole size was larger than 100 mm has
been provided. Previously, it was stated that the hole size was > 100 mm in such cases. The
additional information on large hole sizes has provided confidence in estimation of a more
accurate frequency for large leaks. Secondly, it has been found that the population data (i.e.
number of equipment years) in HCRD has not been updated after 2005, which means that
the estimated frequency for leaks extracted from HCRD will lead to an excessive estimate of
the leak frequency even for installations located on the UKCS
The mathematical formulation in PLOFAM enables an improved representation of the effect
of the equipment size on the hole size frequency distribution for the various equipment
types. In SHLFM, the capability in terms of capturing the shift in hole size distribution with
varying equipment size for a given equipment was less pronounced. In combination with
parameterisation of SHLFM outside the range of HCRD data for large holes led to estimation
of excessive frequency for large holes for all equipment sizes for all equipment type.
Moreover, as additional data on equipment size per incident in HCRD has been made
available to the project, it has been possible to develop and parameterize the equipment size
dependent model in PLOFAM
PLOFAM is unbiased. Bias of frequency towards large hole sizes were included in SHLFM to
account for uncertainty.
The resulting quantitative fire and explosion risk picture in a QRA for a typical installation on the
NCS will be very different based on PLOFAM opposed to SHLFM. The validation model has
demonstrated that the model denoted PLOFAM is able to predict the observed number of leaks
at installations located on the NCS in the period 2001-2014, whilst SHFLM will overpredict the
observed number of leaks in the same period.

Figure 8.1 Comparison of SHLFM and PLOFAM. The frequency is the sum of frequency for gas and
liquid leaks. For the SHLFM model, only Full pressure leaks and Limited leaks are included. In
PLOFAM, only significant leaks are included
Figure 8.2 Comparison of SHLFM and PLOFAM. The bars display ratio per leak category for the
total of gas and liquid leaks. For the SHLFM, only Full pressure leaks and Limited leaks are included.
In PLOFAM, only significant leaks are included

9 MISOF benchmarking model and PLOFAM
The MISOF ignition model (Ref. /3/) was developed in a project run by the Norwegian Oil and Gas
Association (Norsk olje & gass), and reported by LRC in 2014. The benchmark model in the
MISOF ignition model has been rerun to investigate the total effect on the fire and explosion
probability of the combined use of MISOF and PLOFAM.
The results based on the MISOF model combined with SHLFM are shown in Figure 9.1. The
results for the case with the MISOF model combined with PLOFAM are presented in Figure 9.2
(explosion pressure frequency distribution). The results for the OLF ignition model (Ref. /4/)
combined with the leak frequency models have not been investigated. It is expected that the
resulting explosion pressure frequency distribution for the combined use of the OLF ignition
model and the PLOFAM leak frequency model would be in between the distributions displayed in
Figure 9.1 and Figure 9.2.
The results demonstrate the effect of PLOFAM on the fire and explosion pressure is prominent.
The effect will depend on the type of ignition sources constituting the major contribution to the
fire and explosion frequency. For platforms where special sources of ignition, such as gas turbine
air intakes, are a dominating factor, the effect of PLOFAM is more prominent than for other
installations. This is because the exposure probability for special sources of ignition is dominated
by large leaks and the relative reduction in leak frequency using PLOFAM instead of SHLFM
increases with increasing initial leak rate.
Figure 9.1 - MISOF combined with SHLFM: Frequency distribution for explosion pressure for all
modules altogether at Installation A

Figure 9.2 MISOF combined with PLOFAM: Frequency distribution for explosion pressure for all
10 Robustness of PLOFAM
The uncertainty related to the capability of the model to predict the occurrence of future leaks at
installations on the NCS is discussed in Chapter 11 in TN-6.
The following aspects are discussed:
General evaluation of quality of the data basis
The variance in prediction of installation specific leak frequency
Quality of the NCS population dataset
Effect of trend with time and the prediction of future gas leaks
Randomness related to the observation of leaks.
The quality and limitations of the data used as basis for the parameterisation of the PLOFAM
parameters is fundamental for the precision of the model. The quality of the data basis is
discussed throughout the report, both in the technical notes presenting the data basis (TN-2 and
TN-3), but also in discussion of the results generated by the validation model (see TN-6).
It is judged that the elements affecting the quality and limitations of the data are understood, but
some of them may be hard to quantify. On a high level, the frequency distributions based on
data extracted from the HCR database and the NCS database is similar.
It is concluded that PLOFAM capability to predict the installation specific leak frequency is
acceptable. In this regard it must be emphasized that the PLOFAM is equally valid for an
installation where PLOFAM underpredicts the observed number of leaks as for installations where
PLOFAM overpredicts the observed number of leaks. PLOFAM target the average for all
installations located on the NCS.

PLOFAM overpredicts the observed leak frequency at installations NCS for the period after 2007.
This applies for entire spectre of initial leak rates. This follows from the fact that the PLOFAM
parameters have been set based on the entire period 01.01.2001-31.12.2014. The historical
frequency for the period 2001-2014 is considerably higher than for the period 01.01.2007
31.12.2014. The latter period can be expected to be more representative for the future
underlying frequency for leaks at installations on the NCS presumed that the casual factors
leading to leaks in this period is also decisive for the future frequency for leaks (e.g. the effect of
emerging unknown casual factors such as degradation mechanisms due to age or changing
operational conditions is small).
Basing the validation on the period after 2007, it is shown that PLOFAM account for one huge
(initial leak rate > 100 kg/s) gas leak occurring at an installation located on the NCS in the near
future. Assessment of randomness shows that the occurrence of several leaks having an initial
leak rate beyond 10 kg/s in the near future will not lead to rejection of the model.
11 Concluding remarks
The strategy has been to build a model that gives a best estimate for future leak frequencies, i.e.
to create an unbiased model without built in conservatism. It is observed a significant decreasing
trend in historical leak frequency with time for installations on the NCS in the period after year
2000. The period 2001-2014 is used as basis for the model, but the historical frequency for all
installations on NCS the last 5 year period is 40% less than the average for the period 2001-2014.
Hence, presuming that the number of leaks at the NCS in the future will follow the observed
frequency for the last 5 year period, PLOFAM is regarded to give robust results for future average
leak frequencies for installations on the NCS. The period 2001-2014 is used as basis for the
model to account for uncertainty in the data material and shifts in underlying casual factors (e.g.
emerging unknown degradation mechanisms due to age or changing operational conditions)
affecting the future trend in leaks occurring on installations on the NCS.
It must be noted that due to stochastic effects and unique installation specific conditions, the
historical leak frequency per installation at the NCS can vary significantly from the NCS average.
PLOFAM is targeting the average leak frequency for the entire population for the entire period
2001-2014, and a variance around the average value is expected.
and how initial leak rates are modelled in the validation process. The validation model show that
The robustness of the conclusions being made with respect to prediction of future leaks at
installations on the NCS is discussed in TN-6.
The model is thoroughly validated towards available data of leaks that has occurred at
installations both on the United Kingdom Continental Shelf (UKCS) and the Norwegian
Continental Shelf (NCS). A main overall conclusion obtained from running the parameterization
and validation process is that the underlying hole size frequency distribution for equipment at
installations located on the NCS is similar to the distribution for equipment located on UK
installations. The differences may be explained by uncertainty related to the datasets (both the
leaks and the population data), limitations of the mathematical formulations and uncertainty

associated with the parameterization process. A set of objectives for further work has been
suggested by the project to close identified gaps in the fundamental data material used as basis
for the model development (see TN-6).
PLOFAM has been compared with the commonly used leak frequency model denoted SHLFM
(Standardised Hydrocarbon Leak Frequencies Model, Ref /2/). The difference between the leak
frequencies generated by the two models is considerable. These differences are explained by
effects following from properties of the new data material being available for development of
PLOFAM as well as new features of the mathematical framework enabling improved
representation of the equipment size for the various equipment types. The resulting quantitative
fire and explosion risk picture in a QRA for a typical installation on the NCS will be significantly
different based on PLOFAM opposed to SHLFM. The validation model has demonstrated that
PLOFAM is able to predict the observed number of leaks in the period 2001-2014 at installations
located on the NCS, whilst SHFLM will overpredict the observed number of leaks for the same
period.
The benchmark model in the MISOF (Ref. /3/) ignition model has been rerun to investigate the
total effect on the fire and explosion probability of the combined use of MISOF and PLOFAM. The
results demonstrate the effect of PLOFAM on the fire and explosion frequencies is prominent.
fire and explosion frequency. For platforms where special sources of ignition, such as gas turbine
air intakes, are a dominating factor, the effect of PLOFAM relative to the use of SHLFM is more
prominent than for other modules. This is because the exposure probability for special sources of
ignition is dominated by large leaks. The relative reduction in leak frequency using PLOFAM
instead of SHLFM increases with increasing initial leak rate.
PLOFAM is validated towards data gathered from installations at NCS. However, it is
demonstrated that the PLOFAM hole size frequency distributions for equipment at installations
located on the NCS is similar to the distributions for equipment located on installations on the
UKCS. The model is therefore judged to be applicable also for installations on the UKCS.
Although the model is based on releases of hydrocarbons from process equipment on North Sea
offshore facilities, it is found reasonable to argue that the model is applicable to platforms and
land based facilities in other domains. This should be based on a specific assessment to qualify
use of the model in the particular domain. The important element to evaluate is whether the
properties of the equipment and operation conditions can be considered equivalent with what
are found generally on installations located in the North Sea.

12 References
/1/ Lloyds Register Consulting, Blowout and well release frequencies based on SINTEF
offshore blowout database 2014, 17 March 2015, Report No: 19101001-8/2015/R3 Rev:
Final
/2/ DNV, Offshore QRA Standardised Hydrocarbon Leak Frequencies, report number 2009-
1768, rev. 1, 16.01.2009.
/3/ Scandpower AS: Modelling of ignition sources on offshore oil and gas facilities, Report
th
No. 102657/R1, September 18 2014.
/4/ Scandpower AS: "Ignition modelling in risk analysis", report no. 89.390.008/R1, March
2007.

Working together
for a safer world
Abbreviations and expressions

TN-1
Technical note for:

Technical note no: 105586/TN-1 Rev: Final A

Date: 18 March 2016
1 Introduction .....................................................................................................................................1
2 Abbreviations ..................................................................................................................................1
3 Expressions ......................................................................................................................................2
3.1 Full bore hole ........................................................................................................................2
3.2 Probability density function ...................................................................................................2
3.3 Frequency density function ....................................................................................................2
3.4 Cumulative hole size probability distribution ..........................................................................2
3.5 Cumulative hole size frequency distribution ...........................................................................2
3.6 Complementary cumulative hole size probability distribution .................................................3
3.7 Complementary cumulative hole size frequency distribution ..................................................3
3.8 Incident data .........................................................................................................................3
Incident data gives information about specific events where HC-leaks have been detected. .............3
3.9 Population data/Exposure data ..............................................................................................3
Technical note no: 105586/TN-1 Rev: Final A Page ii

1 Introduction
This technical note gives abbreviations used in the main report and TN-2 TN-6. Also the most
important expressions used throughout the report are described in detail.
2 Abbreviations
Table 2.1 gives abbreviations used in the main report and TN-2 TN-6.
Table 2.1 - Abbreviations used in the main report and TN-2 TN-6
ASCV Annulus safety check valve
ASV Annulus safety valve
AWV Annulus wing valve
BD Blowdown
BOP Blowout preventer
BORA Barrier and operational risk analysis
DHSV Downhole safety valve
EQCDB Equipment count database
ESD Emergency shut down
ESV Emergency safety valve
GLV Gas lift valve
HAZID Hazard Identification
HC Hydrocarbon
HSE Health and safety executive
LFL Lower flammability limit
LRC Lloyds Register Consulting
Modelling of ignition sources on offshore oil and gas
MISOF
Facilities
N/A Not applicable
Process Leak for Offshore installations Frequency
PLOFAM
Assessment Model
P&ID Piping and instrumentation diagram/drawing
PSD Process shut down
Ptil Petroleumstilsynet (Petroleum safety authority)
PWV Production wing valve
Technical note no: 105586/TN-1 Rev: Final A Page 1

QRA Quantitative risk analysis
Risikoniv i norsk petroleumsvirksomhet (Risk level in
RNNP
Norwegian petroleum industry)
SHLFM Standardised hydrocarbon leak frequencies model
TN Technical note
UKCS United kingdom continental shelf
WO Work order
3 Expressions
This chapter gives detailed explanations of expressions used in the main report and all technical
notes.
3.1 Full bore hole

Full bore hole is used to describe holes that have the same dimension as the equipment itself.
That can be caused by a rupture, but also leaks from a valve that is left fully opened or is fully
opened during an operation are categorized as a full bore hole leak.
3.2 Probability density function

The probability density function for hole sizes gives the relative likelihood for a hole size to take
on a given value. The probability of the hole size falling within a particular range of values is
given by the integral of the density function over that range, i.e. it is given by the area under the
density function between the lowest and greatest values of the range. The probability density
function is nonnegative everywhere, and its integral over the entire space is equal to one.
3.3 Frequency density function

The frequency density function for hole sizes gives the frequency for a hole size to take on a
given value. The frequency of the hole size falling within a particular range of values is given by
the integral of the density function over that range, i.e. it is given by the area under the density
function between the lowest and greatest values of the range. The frequency density function for
a particular equipment type is nonnegative everywhere, and its integral over the entire space is
equal to the total leak frequency for that equipment type.
3.4 Cumulative hole size probability distribution

This expression denotes the probability of a leak with hole size equal or smaller than d: P(hole
size d). This is not widely used in the report as the modelling is done for the complementary
cumulative hole size probability distribution P(hole size > d).
3.5 Cumulative hole size frequency distribution

This expression denotes the frequency of a leak with hole size equal or smaller than d: F(hole size
d). This is not widely used in the report as the modelling is done for the complementary
cumulative hole size frequency distribution F(hole size > d).

3.6 Complementary cumulative hole size probability distribution
This expression denotes the probability of a leak with hole size larger than d: P(hole size > d). This
expression is for simplicity often referred to as the hole size probability distribution, throughout
the report.
3.7 Complementary cumulative hole size frequency distribution

This expression denotes frequency of a leak with hole size larger than d F(hole size > d). This
expression is throughout the report denoted F, and for simplicity it is referred to as the hole size
frequency distribution.
Note that the complementary cumulative hole size probability distribution for an equipment type
multiplied by the total leak frequency for that equipment type, gives the complementary
cumulative hole size frequency distribution.
3.8 Incident data

Incident data gives information about specific events where HC-leaks have been detected.
3.9 Population data/Exposure data

Both terms population data and exposure data are used to denote the number of
equipment years in the populations. In other words these expressions are used to denote the
number of units (for instance pumps) multiplied with the number of years in operation.

Working together
for a safer world
NCS data
TN-2
Technical note for:


Date: 18 March 2016
1 Introduction .....................................................................................................................................1
2 General description of dataset .........................................................................................................1
2.1 Leaks at NCS .........................................................................................................................1
2.2 Hole size................................................................................................................................1
2.3 Population data NCS .............................................................................................................1
3 Relevant incidents for the modelled leak scenarios ...........................................................................2
4 Distribution of leaks per system .......................................................................................................2
5 Distribution of leaks per leak scenario ..............................................................................................4
6 Distribution of leaks per leak rate per equipment type ...................................................................10
7 Distribution of leaks per leak rate ..................................................................................................20
8 Distribution of leaks per leak rate per fluid type .............................................................................26
9 NCS population dataset .................................................................................................................30
9.1 General ...............................................................................................................................30
9.2 Number of process equipment units ....................................................................................32
9.3 Steel piping .........................................................................................................................36
9.4 Producing wells and gas lift wells ........................................................................................40
9.5 Hose operations ..................................................................................................................42
9.6 Comparison equipment containing gas and liquid ...............................................................44
9.7 Equipment years ..................................................................................................................46
9.8 Assessment of quality ..........................................................................................................48
9.8.1 General ............................................................................................................................ 48
9.8.2 Valves, flanges and instruments ........................................................................................ 48
9.8.3 Modification projects and equipment years ....................................................................... 51
9.8.4 Time in operation ............................................................................................................. 51
10 Time distribution ............................................................................................................................52
10.1 Overall trend leaks ...............................................................................................................52
10.2 Analysis of time trend ..........................................................................................................54
11 NCS data used for validation of model...........................................................................................59
12 References .....................................................................................................................................60
Appendix A Recorded incidents at NCS

Appendix B Email used as reference

1 Introduction
This technical note describes the recorded leaks occurring on installations located on the
Norwegian Continental Shelf (NCS) in the period 01.01.2001 31.12.2014 (14 years) and
corresponding population data. The population data consist of equipment counts gathered from
QRAs from 68 installations, of which 62 have been in operation in the period.
Abbreviations and expressions used in this technical note are described in TN-1 Abbreviations and
expressions.
2 General description of dataset

2.1 Leaks at NCS
The data of recorded leaks at the NCS have been established based on the following data
sources:
1. RNNP dataset collated by Petroleumstilsynet (Ptil) and Safetec
2. Review of accident investigation reports. Accident investigation reports have been available
for the major fraction of the incidents.
The total number of leaks occurring in the period is 222.
Recorded leaks at NCS have all an initial hydrocarbon leak rate of 0.1 kg/s or larger.
2.2 Hole size

It has been attempted to assess the hole size for the incidents based on review of the accident
investigation reports. Adequate information to determine the hole size is only available in a frac-
tion of the cases, and further analysis of the hole sizes has been disregarded. It is judged that a
detailed review of the incident reports would enable assessment of the hole size for about 50 %
of the leaks. The main reason for hole sizes not being available is that it is not a requirement to
report that parameter. It is recommended that the future practice for reporting of incidents inclu-
des reporting of the hole size (preferably both calculation of the hydraulic hole size (based on
pressure and density) as well as measurements of the physical properties of the observed hole (if
possible).
2.3 Population data NCS

Population data has been collated for 68 installations based on equipment counts extracted from
the QRAs for the installations. 6 out of the 68 installations have not been set in operation by
31.12.2014. Hence, population data is available for 62 installations being in operation in the peri-
od 01.01.2001 31.12.2014. The population data set for the 62 installations is hereafter deno-
ted NCS population dataset.
In addition, population data (i.e. equipment counts) for 12 more installations being in operation
on the NCS in the period (01.01.2001 31.12.2014) has been estimated by defining an equi-
valent installation in the NCS population dataset. The equivalent installation has been based on
an overall evaluation of the installation characteristics. 10 out of 222 leaks at NCS are related to
those 12 installations. The population data set for the 74 installations (62 installations from NCS
population data set + 12 installations) is hereafter denoted Full NCS population dataset.
Population data on topside producing wells, producing wells equipped with gas lift and
operations involving use of hoses have been established based on input from the two operators
operating the major fraction of installations on the NCS.

3 Relevant incidents for the modelled leak scenarios
The total number of leaks occurring in the period 01.01.2001 31.12.2014 is 222. After review
of the incidents, it has been concluded that 31 of those incidents are not relevant for the leak
scenarios to be modelled by the model, i.e. they are not process leaks or topside well leaks during
normal operation (see TN-4). Typical properties of disregarded incidents are as follows:
The leak is a release through a vent or a dump line where the rate is not considered to
exceed the design specification for the vent or dump line
The leak is originating from a piece of equipment not being relevant, such as a pipeline or a
riser
The leak is occurring in the well system during a drilling operation.
Out of the remaining 191 leaks (222 - 31), 181 have occurred on either of the 62 installations in
the NCS population dataset. The remaining 10 leaks have occurred on either of the additional
12 installations in the Full NCS population dataset. Detailed information about all 191 relevant
leaks are given in Appendix A.
4 Distribution of leaks per system

The distribution per system is shown in Table 4.1, i.e. the table gives the number of leaks
registered from the different systems. The categorisation of system in this context may not
coincide with general industry practice in terms of categorisation of systems. The used naming is
assumed to be self-explanatory.
The classification of leaks per system is somewhat uncertain and should be interpreted with care,
which is due to limited information available in terms of the system the incident is associated
with. However, the results can be used to conclude that leaks are in many cases associated with
several systems. The typical scenario in cases where several systems are involved is that the leak is
fed from the process system, but the leak point itself is associated to the other system.
Several incidents are related to the well system, but it is only one incident where it is clearly
concluded that the leak source is not related to a leak source in the process system. It is judged
that a more thorough review would conclude that the leak source in several cases is associated
with the well system and not the process system.

Table 4.1 - Distribution of all NCS leaks (191) per system. The question mark indicates that the leak location is associated with uncertainty
System where leak is released from b(0.1-1 kg/s) c(1-10 kg/s) d(10-100 kg/s) e(>100 kg/s) Grand Total
Closed drain 1 1 2
Closed drain / Open drain system 1 1
Closed drain / Process system 1 1
Flare system 2 3 4 1 10
Fuel gas system 4 2 6
Fuel gas system / Diesel system 1 1
Open drain system 1 1
Process system 101 24 5 2 132
Process system / Closed drain 1 1
Process system / Flare system 4 4
Process system / Fuel gas system 1 1
Process system / Gas lift system 1 1
Process system / Open drain 1 1
Process system / Produced water? 1 1
Process system / Seal oil system 1 1
Process system / Storage 1 1 2
Process system / Utility system 1 1
Process system / Well system 13 4 17
Produced water / Sea water / Open drain 1 1
Produced water system 1 1
Unknown 2 2 4
Well system 1 1
Grand Total 134 42 12 3 191

5 Distribution of leaks per leak scenario
The distribution of the leaks per leak scenario (Significant and Marginal, see TN-4) is presented in
Figure 5.1 through Figure 5.7 and Table 5.1 and Table 5.2. Figure 5.1 and Figure 5.3 shows the
results for the installations in the NCS population dataset (181 leaks), whilst Figure 5.2 and Figure
5.4 display the results for the full NCS population dataset (all 191 leaks). Figure 5.5 shows the
relative distribution with respect to the initial leak rate for the two scenarios separately for the full
NCS population dataset (all 191 leaks).
A Marginal leak is a leak where the total mass released to the environment is 10 kg or less. In a
Significant leak scenario, the released amount is beyond 10 kg.
The results show that the fraction Marginal leaks are about 12 % for both datasets. The fraction
is quite constant for all leak categories, i.e. around 10 %. There is not observed any Marginal
leaks having an initial leak rate larger than 100 kg/s. The number of such leaks in the dataset is
few, and the results are sensitive to randomness.
It is likely that several of the scenarios classified as Significant are Marginal, i.e. the actual amount
released to the environment was less than 10 kg. The total amount released is not stated in some
of the accident investigation reports, and investigation reports for some of the incidents are not
available. In all cases where the amount released is unknown, the scenario type is classified as
Significant.
The average fraction of Marginal leaks is about 12 % for all leak rates and all fluid types, and
should be used to estimate the fraction of Marginal leaks from the total number of leaks. It is
expected that future data will provide basis for increasing this figure somewhat. However, the
distribution with respect to equipment type, shown in Figure 5.6 and Figure 5.7, point towards
that the fraction of Marginal leaks is different for the various types of equipment. The results
shows that a high fraction Marginal leaks is associated with leaks from hoses, producing wells,
valves and flanges. Note that only one incident has been recorded for producing wells (i.e. the
result is sensitive to randomness).
Table 5.1 - Relative distribution of leaks (181) per leak scenario for installations in the NCS
population dataset
Distribution per b(0.1-1 c(1-10 d(10-100 e(>100 Grand
leak scenario kg/s) kg/s) kg/s) kg/s) Total
Marginal leak 13 % 8% 8% 0% 12 %
Significant leak 87 % 92 % 92 % 100 % 88 %
Grand Total 100 % 100 % 100 % 100 % 100 %
Table 5.2 - Relative distribution of all NCS leaks (191) per leak scenario
Distribution per b(0.1-1 c(1-10 d(10-100 e(>100 Grand
leak scenario kg/s) kg/s) kg/s) kg/s) Total
Marginal leak 13 % 10 % 8% 0% 12 %
Significant leak 87 % 90 % 92 % 100 % 88 %
Grand Total 100 % 100 % 100 % 100 % 100 %

Figure 5.1 - Distribution of leaks (181) per leak scenario for installations in the NCS population
dataset
Figure 5.2 - Distribution of all NCS leaks (191) per leak scenario

Figure 5.3 - Relative distribution of leaks (181) per leak scenario for installations in the NCS
population dataset
Figure 5.4 - Relative distribution of all NCS leaks (191) per leak scenario. The contribution from
Marginal and Significant leaks sums up to 100%

Figure 5.5 - Relative distribution of all NCS leaks (191) per leak scenario. The contribution from the
initial leak rate categories sums up to 100%

Figure 5.6 - Distribution of all NCS leaks (191) per leak scenario per equipment type

Figure 5.7 - Relative distribution of all NCS leaks (191) per leak scenario per equipment type. NB! Only one event related to producing wells

6 Distribution of leaks per leak rate per equipment type
The distribution of the leaks per equipment type is shown in Table 6.1 and Table 6.2 and
illustrated in Figure 6.2 through Figure 6.6.
The quality of the assessment of equipment type causing the leak is considered to be high. A
thorough review has been performed to relate the correct equipment type to the leak point.
The leaks from vents are releases from vents where the design specification of the vent has been
exceeded.
Hoses in this context are hoses used for temporary operations, such as bleed of from well
annulus to flare system. 6 % of the leaks are related to use of hose in temporary operations. No
leaks have been found that is related to flexible piping. In terms of technical properties, flexible
piping may similar to hoses (however this is not investigated), but a flexible pipe is equipment
that in general is intended for permanent use (actual use may of course only be a fraction of
time).
The results demonstrate that the equipment type valve is the dominant contributor. The equip-
ment type valve cover all types of valves both in terms of how it is operated (manual vs. actuated
valve), functionality (e.g. safety valve vs. control valve) and design (e.g. gate valve vs. ball valve).
Almost 50 % of the leaks are tagged to valves. The second and third biggest contributors are
standard flange and steel pipe. The three mentioned equipment types plus instruments constitute
about 85 % of the leaks at NCS. All of the incidents related to flanges are from standard flanges.
No leaks stemming from compact flanges have been identified.
The distribution is the same for both population datasets (leaks in population of 62 installations
and population of 74 installations). The distribution is also quite unaffected when looking at only
leaks originating from leak sources in the process system (see Table 6.2 vs Table 6.3).
An important element to note when comparing the data with the HCR data for UKCS installa-
tions is a difference with regard to classification of leaks stemming from valves and flanges
associated with an instrument connection. For the NCS leaks it is judged that it is bias towards
logging leaks stemming from a valve or a flange associated with instrument connections to the
valve or flange, and not to the instrument connection as a whole. For illustration, a couple of
examples of instruments including other equipment types such as valves and flanges are shown
in Figure 6.1.
In some of these cases it is not known or unclear (based on the accident investigation report)
whether the leak point is related to a valve or flange attached to the instrument connection or a
piece of equipment in the vicinity of the instrument connection.
The following statement describes how flanges and valves connected to instruments should be
included in the HCRD population database
One Instrument could comprise the instrument itself, plus up to 2 valves, up to 4 flanges, 1
fitting, and associated small bore piping (1"or less). Corrosion coupons should be treated as
flanged, valved connections with instrument (probe & cap) attached.
The statement is ambiguous, which means that there are some uncertainty related to how valves
and flanges associated with instruments are put in the HCR population data, which is important
to account for when assessing the difference in frequency between flanges, valves and
instruments. This is further discussed in Chapter 9.8.
some incidents in HCRD are logged on equipment type steel pipe and reported as left open
and opened as causes. These incidents should most likely have been logged on valves. It is
assessed that this error is most likely not performed on the NCS data.

.
Figure 6.1 - Examples instrument connections

Table 6.1 - Relative distribution of leaks per equipment type for installations in the NCS population dataset (181)
Distribution per equipment type b(0.1-1 kg/s) c(1-10 kg/s) d(10-100 kg/s) e(>100 kg/s) Grand Total
Compressor 1.1 % 0.0 % 0.6 % 0.0 % 1.7 %
Filter 0.6 % 0.6 % 0.0 % 0.0 % 1.1 %
Hose 2.8 % 3.3 % 0.0 % 0.0 % 6.1 %
Instrument 7.7 % 1.7 % 0.0 % 0.0 % 9.4 %
Pig trap 0.6 % 0.0 % 0.0 % 0.0 % 0.6 %
Process vessel 1.1 % 0.0 % 0.0 % 0.0 % 1.1 %
Pump 1.1 % 0.0 % 0.0 % 0.0 % 1.1 %
Standard flange 8.3 % 4.4 % 2.2 % 0.6 % 15.5 %
Steel pipe 11.0 % 0.6 % 1.1 % 0.6 % 13.3 %
Unknown 0.6 % 0.6 % 0.0 % 0.0 % 1.1 %
Valve 35.4 % 8.8 % 2.2 % 0.6 % 47.0 %
Vent 0.0 % 1.7 % 0.6 % 0.0 % 2.2 %
Grand Total 70.2 % 21.5 % 6.6 % 1.7 % 100.0 %

Table 6.2 - Relative distribution of all NCS leaks (191) per equipment type
Compressor 1.0 % 0.0 % 0.5 % 0.0 % 1.6 %
Filter 1.0 % 0.5 % 0.0 % 0.0 % 1.6 %
Hose 2.6 % 3.1 % 0.0 % 0.0 % 5.8 %
Instrument 7.3 % 1.6 % 0.0 % 0.0 % 8.9 %
Pig trap 0.5 % 0.0 % 0.0 % 0.0 % 0.5 %
Process vessel 1.0 % 0.0 % 0.0 % 0.0 % 1.0 %
Producing well 0.0 % 0.5 % 0.0 % 0.0 % 0.5 %
Pump 1.0 % 0.0 % 0.0 % 0.0 % 1.0 %
Standard flange 8.4 % 4.7 % 2.1 % 0.5 % 15.7 %
Steel pipe 11.0 % 1.0 % 1.0 % 0.5 % 13.6 %
Storage tank 0.5 % 0.0 % 0.0 % 0.0 % 0.5 %
Unknown 0.5 % 0.5 % 0.0 % 0.0 % 1.0 %
Valve 35.1 % 8.4 % 2.1 % 0.5 % 46.1 %
Vent 0.0 % 1.6 % 0.5 % 0.0 % 2.1 %
Grand Total 70.2 % 22.0 % 6.3 % 1.6 % 100.0 %

Table 6.3 - Relative distribution of NCS leaks (191) per equipment type where leaks stemming from equipment in open drain, closed drain, flare system,
well system, produced water system and unknown systems have been disregarded (see Table 4.1)
Compressor 1.2 % 0.0 % 0.6 % 0.0 % 1.8 %
Filter 1.2 % 0.6 % 0.0 % 0.0 % 1.8 %
Hose 2.4 % 2.9 % 0.0 % 0.0 % 5.3 %
Instrument 8.2 % 1.8 % 0.0 % 0.0 % 10.0 %
Pig trap 0.6 % 0.0 % 0.0 % 0.0 % 0.6 %
Process vessel 0.6 % 0.0 % 0.0 % 0.0 % 0.6 %
Pump 1.2 % 0.0 % 0.0 % 0.0 % 1.2 %
Standard flange 8.8 % 4.7 % 0.0 % 0.6 % 14.1 %
Steel pipe 11.8 % 1.2 % 1.2 % 0.0 % 14.1 %
Storage tank 0.6 % 0.0 % 0.0 % 0.0 % 0.6 %
Valve 38.8 % 8.2 % 1.8 % 0.6 % 49.4 %
Vent 0.0 % 0.6 % 0.0 % 0.0 % 0.6 %
Grand Total 75.3 % 20.0 % 3.5 % 1.2 % 100.0 %

Figure 6.2 - Distribution of leaks per equipment type leaks on installations in the NCS population dataset (181)

Figure 6.3 - Distribution of all NCS leaks (191) per equipment type

Figure 6.4 - Relative distribution of all leaks per equipment type leaks on installations in the NCS population dataset (181)

Figure 6.5 - Relative distribution of all NCS leaks (191) per equipment type

Figure 6.6 - Relative distribution of all NCS leaks (191) per equipment type where leaks stemming from equipment in open drain, closed drain, flare
system, well system, produced water system and unknown systems have been disregarded (see Table 4.1)

7 Distribution of leaks per leak rate
The distribution of leaks with respect to initial leak rate is shown in Figure 7.1 through Figure
7.11.
The following conclusions can be extracted from the data displayed in the various figures:
Liquid leaks tend to result in slightly larger leaks than gas leaks. The difference between gas
and liquid leaks is becoming more and more prominent with increasing leak rate, and is
significant for leaks with an initial leak rate greater than 100 kg/s. The fraction of liquid leaks
beyond 100 kg/s is five times bigger than for gas leaks. This may be explained by the higher
density of liquids. The overall fraction of leaks above 100 kg/s is about 2 %. The fraction of
gas leaks for such leaks is between 0.5 and 1 %, and slightly less than 5 % for liquid leaks.
Given a gas leak and a liquid leak having an initial leak rate above 0.1 kg/s, it is about 7
times more likely that the initial leak rate of the liquid leak is larger than 100 kg/s than the
gas leak
The fraction of leaks having an initial leak rate larger than 10 kg/s is about 7 % for both gas
and liquid leaks
The fraction of leaks having an initial leak rate larger than 1 kg/s is about 22 % for both gas
and liquid leaks
The relative leak rate distribution is largely the same for Significant and Marginal leaks. No
leaks having an initial leak rate greater than 100 kg/s has been observed for Marginal leaks.
That is however judged to be rare event, and it is concluded that randomness can explain
why such a leak has not been observed. The duration of such an event would be very small
(~0.1 second)
The relative distribution vs leak rate is the same for leaks in the NCS population dataset (181
leaks) and the full NCS population dataset (191 leaks)
The relative distribution for the dominant equipment types (valves, flanges, instrument and
steel pipe) are different. Valves and instrument tend to produce considerable smaller leak
rates than flanges. The properties of leaks stemming from steel pipe (and hoses, see next
bullet point) in terms of leak rate appear to be somewhat different than leaks from the other
components. Leaks from steel pipes tend to either result in very large leaks (> 100 kg/s),
large leaks (> 10 kg/s) or small leaks (< 1 kg/s). It is registered few intermediate leaks (1 10
kg/s). An explanation that steel pipe leaks is more likely to be originating from full bore leaks
leading to big leaks (e.g. caused by over pressurisation), which partly may explain the high
fraction of large leaks. Both of the large leaks originating from steel pipe having a leak rate
larger than 100 kg/s is caused by over pressurisation. Moreover, Figure 7.10 demonstrates
that the relative distribution for leaks originating from steel pipes is very different for gas and
liquid leaks. For the other equipment types, the relative distribution is quite similar for gas
and liquid leaks. However, the assessment of the difference between equipment types for
the various fluid types is due to the low number of incidents sensitive to randomness (i.e. the
observed result may be due to randomness and do not represent actual underlying causes)
There are not registered leaks from hoses with an initial leak rate above 10 kg/s. The fraction
of leaks (11 in total for all leaks at NCS) is split approximately evenly between leaks in the
intervals 0.1 1 kg/s and 1 10 kg/s. The distribution is clearly different from other
equipment types (see Figure 7.11). This is caused by the large rupture fraction of hoses, and
that the largest temporary hose size is 2
No leaks having a leak rate larger than 10 kg/s have originated from a different equipment
type than valve, standard flange, instrument or steel pipe.

Figure 7.1 - Distribution of leaks (181) per leak category for installations in the NCS population
dataset
Figure 7.2 - Distribution of all NCS leaks (191) per leak category

Figure 7.3 - Relative distribution of leaks (181) per leak category for installations in the NCS
population dataset
Figure 7.4 - Relative distribution of all NCS leaks (191) per leak category

Figure 7.5 - Complementary cumulative probability distribution of NCS leaks (191) with respect to
initial leak rate
Figure 7.6 - Complementary cumulative probability distribution with respect to initial leak rate for
NCS leaks (191) and for leaks on installations in the NCS population dataset

leaks on installations in the NCS population dataset
leaks on installations in the NCS population dataset (181)

NCS leaks (191)
NCS leaks (191) per fluid phase

Figure 7.11 - Relative distribution of leaks from hoses with respect to initial leak rate for all NCS
leaks (191) per fluid phase
8 Distribution of leaks per leak rate per fluid type

The distribution of leaks with respect to initial leak rate is shown in Table 8.1, Table 8.2 and
Figure 8.1 through Figure 8.5.
Leaks categorized as condensate, oil or well fluid is considered to be liquid leaks when applying
the data for validation of the model in TN-6. Furthermore, leaks with unknown fluid is distributed
according to the overall relative distribution for the known incidents (see Chapter 11). The resul-
ting distribution per fluid phase (gas or liquid) is shown in Figure 8.5.
The main conclusion extracted from the data is that about 3 out 4 leaks are gas leaks. This
fraction is quite independent of initial leak rate except for the very large leaks (having an initial
leak rate beyond 100 kg/s). For leaks with rate above 100 kg/s, the fraction of liquid leaks is very
different. For such leaks, 2 out 3 is liquid leaks. This may be explained by the higher density of
liquid leaks (i.e. the leak rate is considerable higher for the same hole size for liquid leaks than for
gas leaks) and/or unique causes for leaks related to steel pipe (see Figure 7.10 and statement
evaluation in previous section). But the number of leaks in this category is only 3, and the relative
distribution as well as the assessment of the causes is sensitive to randomness.

Table 8.1 - Distribution of leaks per fluid type for installations in the NCS population dataset (181)
Distribution fluid type Condensate Gas Oil Unknown Well fluid Grand Total
b(0.1-1 kg/s) 5 98 18 5 1 127
c(1-10 kg/s) 1 28 5 3 2 39
d(10-100 kg/s) 9 3 12
e(>100 kg/s) 1 1 1 3
Grand Total 7 136 27 8 3 181
Table 8.2 - Distribution of all NCS leaks (191) per fluid type
Distribution fluid type Condensate Gas Oil Unknown Well fluid Grand Total
b(0.1-1 kg/s) 5 101 22 5 1 134
c(1-10 kg/s) 1 29 6 3 3 42
d(10-100 kg/s) 9 3 12
e(>100 kg/s) 1 1 1 3
Grand Total 7 140 32 8 4 191
Figure 8.1 - Distribution of leaks per fluid type for installations in the NCS population dataset (181)

Figure 8.2 - Distribution of all NCS leaks (191) per fluid type
Figure 8.3 - Distribution of NCS leaks per fluid phase

Figure 8.4 - Distribution of all NCS leaks (191) per fluid
Figure 8.5 - Distribution of NCS leaks per fluid phase for installations in the NCS population dataset
(181)

9 NCS population dataset
9.1 General
Population data that corresponds to the recorded leaks in period 01.01.2001 31.12.2014 has
been collated for 62 installations based on equipment counts extracted from the QRAs for the
installations. 181 of the total number of leaks at NCS (191) did occur at either of those 62
installations. In addition to the 62 installations, similar equipment counts are available for 6
installations that are put on stream after 31.12.2014. The equipment counts from these 6
installations have not been used in the validation model.
The population dataset consisting of the 62 installations that has been on stream in the period is
denoted NCS population dataset. Counts of steel pipe length are available for 28 installations
out of the 68 installations, of which 24 has been in operation in the period 01.01.2001
31.12.2014.
Population data (i.e. equipment counts) for 12 more installations being in operation on the NCS
in the period (01.01.2001 31.12.2014) has been estimated by defining an equivalent installa-
tion taken from the NCS population dataset. The equivalent installation has been based on an
overall evaluation of the installation characteristics. The population data set for the 74 installa-
tions is denoted Full NCS population dataset.10 out of the 191 leaks at NCS are related to
those 12 installations, which means that the complete population dataset cover all 191 leaks
occurring in the period studied.
The names of the installations are anonymized throughout the report. The age of the installations
as wells as the years in production in the period 01.01.2001 31.12.2014 are not reported to
ensure that the installations cannot be identified.
The six installations not being in operation in the period and corresponding total number of eq-
uipment are presented in Table 9.1 (steel pipe length, hose operations and producing wells are
excluded). The average number of equipment counts is about the same for those 6 installations
and the 62 producing installations.
The estimated equipment count for the 12 installations not in the NCS population dataset is
shown in Table 9.2 (Platform 74 has been disregarded as it is an accommodation platform, but is
kept in the table to show why Platform 74 is not found in the figures throughout the report). The
average number of equipment count for those 12 installations are significantly less than for the
average for the 62 installations. This is mainly because 4 out of those 12 installations are small
wellhead platforms. Note that steel pipes, hose operations and producing wells are included in
the figures and in the table.
Population data on hose operations and producing wells are established specifically in the project
based on information received from the two operators with the most installations on NCS.

Table 9.1 - Platforms in NCS population dataset not commissioned by 31.12.2014. Steel pipes, hose
operations and producing wells are disregarded
Platform not commissioned by 31.12.2014 Number of equipment

Platform 13 4 816
Platform 34 4 894
Platform 35 1 449
Platform 39 1 210
Platform 51 4 768
Platform 64 5 309
Grand Total 22 445
Average 6 installations not commissioned by 31.12.2014 3 741
Average 62 installations commissioned by 31.12.2014 3 646
Table 9.2 - Installations on NCS with registered leaks in period 01.01.2001 - 31.12.2014 not included
in NCS population dataset, and equivalent platform taken from NCS population dataset based on
assessment of general installation properties. The total population dataset is denoted Full NCS
population dataset. Steel pipe, hose operations, producing wells is disregarded
Installations not in NCS Equivalent installation Number of equipment

population dataset taken from NCS population
dataset
Platform 69 Platform 53 2 396
Platform 70 Platform 24 280
Platform 74 (Accommodation
N/A N/A
platform)
Grand Total 23 913
Average 1 993
Average 62 installations commissioned by 31.12.2014 3 646

9.2 Number of process equipment units
Figure 9.1 through Figure 9.4 shows the overall equipment counts for the dominant types of
equipment; namely valves, standard flanges and instruments. The contribution from steel piping
is not included.
Standard flanges include all types of flanges. The number of special types of flanges, such as SPO
compact flanges and graylock, are in any case judged to be negligible relative to the number of
standard ASME flanges. The model for compact flanges is based on a separate evaluation of
available population data on compact flanges, which is presented in TN-6. No model has been
developed for graylock flanges.
More than 95 % of the equipment in terms of number of units is either a valve, flange or
instrument.
Of the remaining equipment, vessels and heat exchangers are the dominant contributors.
The equipment type air-cooled heat exchanger registered in the HCR database has not been
registered in the equipment counts.
The count of atmospheric vessels are only two in the whole population, which means that such
equipment units and the systems such equipment units are associated with are generally not
included in QRAs.
The equipment types air-cooled heat exchangers and atmospheric vessels can therefore not be
validated with the validation model.

Figure 9.1 - Distribution equipment counts for dominant type of equipment (standard flange, valve
and instrument). Steel pipe is not included as part of the category Others

Figure 9.2 - Distribution equipment counts for compressors, pumps, heat exchangers, vessels, filters
and pig traps

Figure 9.3 - Relative distribution equipment counts for dominant type of equipment (standard
flange, valve and instrument). Steel pipe is not included as part of the category Others

Figure 9.4 - Over relative distribution equipment counts for compressors, pumps, heat exchangers,
vessels, filters and pig traps
9.3 Steel piping

The count of steel pipe length per platform (24 installations) being in operation in the period
01.01.2001 31.12.2014 is shown in Figure 9.5.
The equipment counts are in general based on either of the following methodologies:
a) Measurement based on iso-view drawings of piping layout
b) Assessment of pipe length based on general equipment layout and platform layout.
The accuracy of method a) is considered significantly better than method b).
Figure 9.7 displays the ratio of steel pipe length divided by number of valves. The results
demonstrate that there is a considerable variance. The number of counts of valves is generally
judged to be the most accurate number. It is also reasonable to expect that the ratio should not
vary too much. It is hard to assess whether the variation represent actual difference in design or
are a result observed due to inaccurate counting. Furthermore, it is difficult argue whether the
counts of steel pipe length generally represents an overprediction or underprediction relative to
actual design.
Hence it is judged that the use of the counts of steel pipe length in the validation model must
account for some uncertainty when concluding on steel pipe leak frequency.

Figure 9.5 - Steel pipe per platform for installations where counts of steel piping is included

Figure 9.6 - Number of valves and steel pipe for installations where counts of steel piping is
included in the NCS population dataset

Figure 9.7 - Ratio steel pipe length divided by number of valves for installations where counts of
steel piping are included in the NCS population dataset

9.4 Producing wells and gas lift wells
The number of wells at NCS is estimated for Statoils and ConocoPhillips installations, by the
respective operators. For Statoils installations, only the total number of well are given. They are
assumed to be distributed homogenously among Statoils installation, resulting in equal number
of wells at different installations.
The number of topside wells and wells with gas lift for the various installations per platform in
the NCS population dataset are shown Figure 9.8. The wells on the additional installations in the
full NCS population dataset are shown in Figure 9.9. The total number of wells is summarized in
Table 9.3.
Topside wells are the total of gas injection wells and producing wells (both wells with and
without gas lift are included). Wells with gas lift are the producing wells equipped with gas lift.
The uncertainty related to the estimate of the number of wells is considered to be small.
Table 9.3 - Topside wells

Row Labels Total number of topside Wells with gas lift
wells
NCS population dataset 756 314
Additional in full NCS population dataset 99 33
Grand Total 855 347

Figure 9.8 - The total number of topside producing wells (wells with and without gas lift) and the
number of producing wells with gas lift on the installations in the NCS population dataset (62
installations)

Figure 9.9 - The total number of producing wells (wells with and without gas lift) and the number
of producing wells with gas lift on the 12 installations added to the NCS population dataset to
obtain the full NCS population dataset
9.5 Hose operations

Hose operations are defined as operations involving temporary use of hose(s). Examples of such
operations are:
Bleed off of well annulus to flare system
Washing/flushing of systems.
The number of temporary operations involving use of hose for 5 installations in the NCS popula-
tion dataset is shown in Figure 9.10. The average number of hose operations for these 5 installa-
tions are 170.
In addition to data from these 5 installations, data from three more installations are available
(Ref. /1/). The average number of hose operations per annum for those three installations are 214
(average of 260, 371 and 12). The average in this case is weighted with the activity level on the
various installations based on the total number of work orders on pressurized equipment. Work
orders are gathered for the three installations for one year in the period (2001 - 2014).
The data demonstrates that the number of operations involving use of hoses varies considerably
in the population. The use of hoses will depend on the design of the well and process systems. It
is also judged that it will depend on the type of installation. It is expected that platforms with
topside wells will have more hose operations than platforms with subsea wells.
The uncertainty associated with the estimate is significant, which must be accounted for when
validating the leak frequency associated with hose operations. Both the overall average and the
median for the two data sets presented above become about 191. It is concluded that using 150
hose operations per installation per year will adequately account for uncertainty when validating
the model for hose operations towards the total number of observed leaks from hoses in the full

NCS population data set (i.e. 11 leaks). The actual average number of hose operations is judged
to rather be 200. Using 150 instead of 200 implies that the estimated frequency for leaks
stemming from hoses becomes 25 % higher than the best estimate.
The distribution of the size of the hoses used is shown in Figure 9.11. It is expected that most
operations are performed with pressurized gas. Hence, the difference with respect to fluid phase.
The assumed density and pressure under the operations are presented in Table 9.4, which are
based on typical conditions under hose operations.
Table 9.4 - Pressure and density in hose operations for various fluid phases
Fluid phase Pressure (bara) Density (kg/m3)
Gas 200 200
Liquid 10 800
Figure 9.10 - Hose operations per year. The green pole is the average of the 5 installations

Figure 9.11 - Distribution hose size in NCS population dataset (G = fraction of all operations with
pressurized gas, L = fraction of operations with pressurized liquid)
9.6 Comparison equipment containing gas and liquid

The distribution of equipment size for the main components, split on equipment containing gas
and liquid are presented in Figure 9.12 and Figure 9.13.
The following can be extracted from the data:
The relative distribution is about the same for both gas and liquid equipment
The number of units containing liquid is considerable less than the number of equipment
containing gas.
This means that the total number of leaks above a certain hole size are to be dominated by
equipment containing gas. In terms of distribution with respect to initial leak rate, this may not
be the case for the population as also the operating pressure and density affects the distribution.

Figure 9.12 - Accumulated distribution number of equipment versus equipment dimension,
equipment containing gas and liquid separately, for all 68 installations in NCS population dataset
Figure 9.13 - Complementary cumulative probability distribution number of equipment versus

equipment dimension, equipment containing gas and liquid separately, for all 68 installations in
NCS population dataset

9.7 Equipment years
The total number of equipment years per year is presented in Figure 9.14 through Figure 9.16.
Figure 9.17 presents the total number of equipment years per equipment type in NCS.
The equipment years are calculated per year by multiplication of the fraction of the year the plat-
form has been in operation. The fraction of the year the platform has been in operation does
only reflect the commissioning and decommissioning date. Accordingly temporary installation
shut downs due to turn around or modification projects have not been reflected.
Based on information provided by the service providers and a brief review of the equipment
counts received, equipment only used (i.e. pressurized) part of the time are to some extent reflec-
ted. A typical example is pig launchers and pig receivers. In any case, equipment being
pressurized temporarily is expected to have negligible effect to the number of equipment years.
The aggregated number of installation years in the two population datasets is shown in Table 9.5.
The additional number of installation years in the full NCS population dataset is about 17 %,
whilst the additional number of equipment years is 8 %. The reason is that the additional
installations in the full NCS population dataset are smaller than the average installations in the
NCS population dataset.
Table 9.5 - Installation years in population datasets in period 01.01.2001-31.12.2014
Population dataset Installation years 01.01.2001-31.12.2014

NCS population dataset 802
Full NCS population dataset 922.9
Figure 9.14 - Equipment years NCS population dataset (corresponding to 181 leaks)

Figure 9.15 - Equipment years full NCS population dataset (corresponding to 191 leaks)
Figure 9.16 - Equipment years in NCS population dataset and additional number of equipment
years estimated for the 12 installations added to obtain the full NCS population dataset

Figure 9.17 - Equipment years per equipment type in NCS population dataset in the period 2001-
2014 (corresponding to 181 leaks). The y-axis is in logarithmic scale
9.8 Assessment of quality

9.8.1 General
The quality of the equipment counts are considered to be high. The data are gathered from QRAs
performed by 4 different service providers. In general, the variation between the different service
providers is small, which indicates that the industry practice on counting of equipment is quite
homogenous. This is interpreted as an argument why the quality of the population data is
regarded as good.
This quality of the counts of producing wells is also considered to be high.
There is significant uncertainty related to steel piping and hose operations, which is addressed in
section 9.3 and section 9.5.
The following aspects related to the quality of the population data is discussed in this section
Relative distribution between valves, flanges and instruments
Modification projects.
9.8.2 Valves, flanges and instruments
The main difference between the various service providers (consultancies delivering QRAs to the
industry in Norway) is a slight variation with regard to the relative distribution of valves, flanges
and instruments (results are not shown). However, the difference between the various services
providers are small relative to the difference between the overall results from the NCS population
dataset compared with the HCR population data (see Figure 9.18 and Figure 9.19. According to
the results, the number of flanges per valve is somewhat less in the UKCS population data
compared to the NCS population data.
In HCRD, a flanged joint is counted as two flanges, opposed to the equipment counts in QRAs in
the industry in Norway, where a flanged joint is counted as one flange. In order to compare the
data with the NCS population data, the number of flanges in HCRD is divided by a factor of 2.

That will lead to a somewhat underestimation of the number of flanges in UKCS installations
because also the contribution single flanges are divided by a factor of two. This is interpreted as
the main reason for the observed deviation, even though other reasons are given below.
The observed difference between the two datasets can to some extent be explained by difference
in the practice on how instruments have been logged. The practice on how flanges and valves
associated with instruments are counted in the industry in Norway may be slightly different from
how instruments are logged as basis for the HCRD population data (see Chapter 6). In total it is
hard to evaluate the total effect of inconsistency in terms of counting of instruments.
A hypothesis can be that fewer flanges per valves at UKCS are due that flanges associated with
instruments are in general represented by the counted instrument in HCRD, and hence left out of
the equipment count.
On the other hand, the NCS population dataset may be affected by that the counting
methodology in general tends to only rely on flanged connections displayed on P&IDs. Generally,
P&IDs does not indicate all flanged connections, particularly flanged small units. This means that
the number of flanged connections may be somewhat under predicted in the NCS population
data. This underestimation is expected also in future QRAs as the counting will still be based on
P&IDs.
The difference display may also represent an actual difference in design, i.e. the use of flanged
connections. Installations on UKCS are in general older than installations on NCS. The application
of welded connections have increased over the recent years, and it is found reasonable that
flanged connections (on average) are more common at UKCS that at NCS. This points towards an
even greater difference between NCS and UKCS in terms of number of flanges relative to the
number of valves and instruments.
The uncertainty related to the counting of valves is judged to less, but may also be affected by
inconsistency related to counting of valves associated with instruments.
It is hard to conclude on the total effect of the above, but the most likely situation is that there is
counted slightly too few standard flanges in the NCS population dataset, which appears to be
the case for the data in HCRD from UKCS as well. The general effect of lack of equipment in the
population data is that the failure frequency per component will be overestimated (the denomi-
nator is too small).
The overall uncertainty related to the population data is however considered small. Broadly, the
relative distributions for NCS and UKCS are similar, which is interpreted as an argument for the
good quality of both datasets. It is concluded that the total number of equipment counts for
instruments, flanges and valves are very reliable. Hence, when validating the model (see TN-6),
the contribution from instruments, flanges and valves should be considered altogether as well as
separately. In other words, a moderate variation in the estimated contribution from valves,
flanges and instruments can be accepted as long as the total contribution from these
components relative to the observed number is acceptable.

Figure 9.18 - Overall ratio number of flanges divided by number of valves for various equipment
size categories for NCS population dataset vs. UKCS HCRD (population data period 1992-2015)
Figure 9.19 - Relative distribution instruments, valves and flanges NCS population dataset vs. UKCS
(HCRD population data period 1992-2015)

9.8.3 Modification projects and equipment years
Major modifications to the installations in the period 2001-2014 are not systematically handled in
the NCS population dataset. A major modification in this context is installation of a new process
module or a compression module.
The equipment counts from the most recent QRA have been used as basis for establishment of
the population data. For installations that have been subjected to major modifications, it is expec-
ted that the recent QRA in most cases includes the equipment counts related to the modification
project. The operational time of the new equipment installed in the period is less than for the
other equipment on the installation. As this is not reflected, the number of equipment years may
be somewhat overestimated for a few installations. On average, this is not judged to affect the
total number of equipment years very much. A conservative assessment of the effect is to assume
that 10 % has been added to 10 % of the installations in the period. Assuming that all this
equipment was set in operation in later half of the last year of the period (in 2014) on installa-
tions that was commissioned before 2001 (and that the average number of equipment on these
installations equals the number of equipment on the average installation in the NCS population
dataset) gives a total effect on the number of equipment years of 1 % (10 % times 10 %). The
actual effect is less as there also will be cases where the equipment counts are not updated with
modifications that have been in operation for a significant time in the observation period. Hence,
it can be concluded that modification projects adds marginal uncertainty to the NCS population
dataset and the validation model in TN-6.
More important than modification projects are the contribution from installations not in the NCS
population dataset. The additional number of equipment years in the full dataset constitutes
about 8 % relative to the number of equipment years in the NCS dataset. The validation model
should therefore consider both the NCS population dataset and full NCS population dataset. The
sensitivity to modifications on the number of equipment years in the NCS population dataset is
discussed further in Chapter 11.4.1 in TN-6.
9.8.4 Time in operation
The equipment counts are gathered from QRAs performed by 4 different service providers. The
procedures for how to take into account that equipment is not in operation the entire year, may
differ across these companies. One common way of doing this is to adjust the quantity parameter,
i.e. the number of equipment in the equipment counts. In the gathered equipment counts it is
seen that some equipment is registered with a quantity different from an integer number. It is
likely that this indicates that adjustments have been made to take reduced operational time into
account.
To get an impression of the uncertainty this may give, pig traps have been analysed in more
detail. As all pig traps are in operation only a small fraction of the full operational time, all pig
traps should have been adjusted. About 30% of the registered pig traps are registered with a
quantity less than 1. Thus, this check may indicate that only about 30% of the equipment counts
have been adjusted, i.e. the equipment counting database is dominated by equipment counts
that are not adjusted. Hence, there is inconsistency in the way this is done in the equipment
counts collected from the different service providers. The potential effect of the inconsistency in
operational time in the population database is discussed further in TN-6 (Chapter 11.4).

10 Time distribution
10.1 Overall trend leaks
The occurrence of leaks versus time is displayed in Figure 10.1 through Figure 10.4. The results
demonstrate that there is a clear trend with time. The number of leaks per year having an initial
leak rate of 0.1 kg/s or greater has decreased considerably.
The trend is not apparent for installations that has been set in operation in the period, but cannot
be assessed by just studying the occurrence of leaks per year as the number of installation set in
operation per year affects the results to a large degree.
Figure 10.1 and Figure 10.2 show that there is no obvious trend for leaks greater than 10 kg/s.
In order to fully understand the trend with time, also the trend in equipment years should be
taken into account. In the following chapter, the combined effect of occurrence of leaks and
equipment years is studied.
Figure 10.1 - Distribution of leaks ( 0.1 kg/s) versus time for installations in the NCS population
dataset (181)

Figure 10.2 - Distribution of NCS leaks (191) versus time (leak rate 0.1 kg/s)
Figure 10.3 - Distribution of NCS leaks (178) versus time for installations commissioned before
01.01.2001

Figure 10.4 - Distribution of NCS leaks (13) versus time for installations commissioned in the period
01.01.2001 31.12.2014
10.2 Analysis of time trend

In Figure 10.5 through Figure 10.12, the average frequencies per component in the NCS
population datasets are presented. The figures are obtained by dividing the number of leaks per
year with the number of equipment years per year. Results are shown for two different subsets in
terms of date of commissioning relative to first day of the period studied (01.01.2014). This is to
evaluate whether the time trend is explained by a difference between new installations set in
operation in the period 01.01.2001 31.12.2014, and older installations set in operation prior to
01.01.2014.
In addition to the results per year, which displays the variation in observed frequency throughout
the period, the average frequency for the entire period plus 5 year average floating average is
shown. Note that the floating average is plotted for the middle year in the 5 year period, i.e. the
5 year average for year 2012 is the average for the period 01.01.2010 31.12.2014.
The following are extracted from the results:

-5
The average leak frequency per component is around 610 per year for the entire period for
-6
all leaks and 610 per year for leaks 10 kg/s. The average for the last 5 years of the period
for leaks having an initial leak rate >0.1 kg/s, is about 40 % less than the average for the
entire period. For leaks having an initial leak rate >10 kg/s the corresponding number is
about 10%. The average for the recent three years is even less. This means a model that is
benchmarked towards the average for the entire period will result in a model that are very
likely to overpredict the future average leak frequency for installations on the Norwegian
Continental Shelf. This aspect is taken into account when the model is parameterized and
validated (see TN-6)
The trend with time is similar for the NCS population dataset and the full NCS population
dataset

Despite the large difference in the variation in leak frequency per equipment per year for
each particular year, the difference in average frequency between installations commissioned
before and after 2001 is small. The 5 year floating average reveals that the average for the
subsets (commission date relative to 01.01.2001) is quite similar the recent years. The
difference is attributed to randomness without any further statistical analysis of the
significance of the difference. The subsets are therefore pooled in the validation model
presented in TN-6
Appear to be a small trend for releases > 10 kg/s (10 %).
Figure 10.5 - Average frequency for leaks ( 0.1 kg/s) per component versus time for installations in
the NCS population dataset (181)
Figure 10.6 - Average frequency and floating average for leaks ( 0.1 kg/s) per component versus
time for installations in the NCS population dataset (181). Floating average plotted versus the
middle year of the period that is averaged (i.e. the data point for year 2012 applies for the period
2010-2014)

Figure 10.7 - Average frequency and floating average for leaks ( 10 kg/s) per component versus
time for installations in the NCS population dataset (181). Floating average plotted versus the
middle year of the period that is averaged (i.e. the data point for year 2012 applies for the period
2010-2014)
Figure 10.8 - Average frequency for leaks ( 0.1 kg/s) per component versus time for all NCS leaks
(191) combined with the full NCS population dataset

Figure 10.9 - Average frequency and floating average for leaks ( 0.1 kg/s) per component versus
time for all NCS leaks (191) combined with full NCS population dataset. Floating average plotted
versus the middle year of the period that is averaged (i.e. the data point for year 2012 applies for
the period 2010-2014)
Figure 10.10 - Average frequency and floating average for leaks ( 10 kg/s) per component versus
time for all NCS leaks (191) combined with the full NCS population dataset. Floating average
plotted versus the middle year of the period that is averaged (i.e. the data point for year 2012
applies for the period 2010-2014)

Figure 10.11 - Average frequency for leaks ( 0.1 kg/s) per component for various time periods and
the two NCS population datasets
Figure 10.12 - Average frequency for leaks ( 10 kg/s) per component for various time periods and
the two NCS population datasets

11 NCS data used for validation of model
The historical leaks recorded on installations on NCS used as basis for validation of the model
described in TN-6 is enclosed in Appendix A.
There are two assumptions introduced that should be kept in mind when evaluating the
validation model;
The equipment type causing the leak is unknown for two incidents. Those two incidents are
not included when benchmarking the leak frequency for the various types of incidents, but
accounted for when benchmarking the overall frequency (see TN-6)
It is 8 incidents where the fluid phase is unknown (see Chapter 8). 6 out of these incidents
have a release rate less than 1 kg/s. The last two incidents have a release rate between 1 and
5 kg/s. The fluid phase in these cases is set randomly according to the overall distribution of
leaks with respect to fluid phase. Hence, 6 out of the leaks are assumed to be gaseous. Thus
the fluid phase for two of the leaks is assumed to be liquid. The effect of this assumption on
the validation model is marginal.

12 References
th
/1/ Email from Jon Andreas Hestad (Safetec) received September 25 2015.

Appendix A
Recorded incidents at NCS
Report no: 105586/TN-2 Rev: Final A

1 Introduction .................................................................................................................................. A1
2 Recorded incidents at NCS ............................................................................................................ A2
Report no: 105586/TN-2 Rev: Final A Page Ai

1 Introduction
This appendix is a part of TN-2. It lists every relevant incidents recorded at NCS, with initial leak
rate 0.1 kg/s, in the period 01.01.2001 31.12.2014. In total 222 incidents are recorded. They
are given a unique ID ranging from 1 222. 191 of the 222 incidents are regarded as relevant,
and are listed in Table 2.1. Detailed description of the data fields in Table 2.1 are given in
Table 1.1.
Table 1.1 - Detailed description of the data fields in Table 2.1
Heading Description
ID ID running from 1 to 222.
Year The year that the leak occurred
Installation Anonymized name of the installation
In NCS population dataset Yes, if the leak occurred at an installation that
is included in the NCS population data set. No
otherwise. See TN-2
Initial leak rate 2015 [kg/s] Initial leak rate based on a thorough review of
investigation reports performed by LRC and
Safetec in 2015.
Medium G=Gas, L = Liquid
Equipment type The equipment type associated with the leak
Leak scenario Leak scenario according to PLOFAM (see TN-4)
Commissioned before 01.01.2001 Yes if the leak occurred at an installation
commissioned before 01.01.2001. No otherwise
Decommissioned before 31.12.2014 Yes if the leak occurred at an installation
decommissioned before 31.012.2014. No
otherwise
System The system associated with the leak.
Report no: 105586/TN-2 Rev: Final A Page A1

2 Recorded incidents at NCS
Table 2.1 All relevant incidents recorded at NCS, with initial leak rate 0.1 kg/s, in the period 01.01.2001 31.12.2014. In total 222 incidents are recorded.
They are given a unique ID ranging from 1 222. 191 of the 222 incidents are regarded as relevant, and are listed in this table
ID Year Installation In NCS Initial Medium Equipment Leak scenario Commissioned Decommissioned System
population leak rate type (see TN-4) before before
dataset 2015 01.01.2001 31.12.2014
[kg/s]
3 2001 Platform 57 YES 0.2 L Valve Significant leak YES NO Process system
4 2001 Platform 55 YES 0.2 G Valve Significant leak YES NO Process system
10 2001 Platform 56 YES 5 G Valve Significant leak YES NO Open drain system
12 2001 Platform 22 YES 0.5 G Standard flange Significant leak YES NO Fuel gas system
14 2001 Platform 21 YES 1.5 G Hose Significant leak YES NO Process system
15 2001 Platform 2 YES 1 G Instrument Significant leak YES NO Process system
18 2001 Platform 53 YES 1.5 G Instrument Significant leak YES NO Process system
19 2001 Platform 7 YES 0.6 G Steel pipe Significant leak YES NO Flare system
20 2001 Platform 51 YES 0.9 L Hose Significant leak YES NO Unknown
21 2001 Platform 23 YES 1.6 G Unknown Significant leak YES NO Unknown

dataset 2015 01.01.2001 31.12.2014
[kg/s]
22 2001 Platform 45 YES 4.7 G Hose Significant leak NO NO Unknown
23 2001 Platform 42 YES 0.1 G Unknown Significant leak NO NO Unknown
24 2002 Platform 55 YES 0.2 G Steel pipe Significant leak YES NO Process system
26 2002 Platform 22 YES 0.2 G Valve Marginal leak YES NO Process system
27 2002 Platform 22 YES 2.5 G Vent Significant leak YES NO Fuel gas system
33 2002 Platform 29 YES 22 G Valve Significant leak YES NO Flare system
34 2002 Platform 54 YES 2 G Vent Significant leak NO NO Produced water
system
36 2002 Platform 4 YES 0.36 G Standard flange Significant leak YES NO Process system
40 2002 Platform 57 YES 10 G Valve Significant leak YES NO Process system
42 2002 Platform 2 YES 0.15 G Standard flange Marginal leak YES NO Process system

dataset 2015 01.01.2001 31.12.2014
[kg/s]
49 2002 Platform 4 YES 0.4 L Valve Marginal leak YES NO Process system /
Well system
60 2003 Platform 7 YES 0.3 L Standard flange Marginal leak YES NO Process system
63 2003 Platform 51 YES 0.1 L Standard flange Marginal leak YES NO Process system /
Well system
64 2003 Platform 27 YES 0.34 L Standard flange Significant leak YES NO Process system
67 2003 Platform 44 YES 9.5 L Valve Significant leak NO NO Flare system
68 2003 Platform 8 YES 0.5 G Valve Significant leak YES NO Closed drain
70 2003 Platform 7 YES 1 G Standard flange Significant leak YES NO Process system /
Fuel gas system
72 2003 Platform 48 YES 0.2 G Filter Significant leak YES NO Process system

dataset 2015 01.01.2001 31.12.2014
[kg/s]
74 2003 Platform 7 YES 1 G Valve Marginal leak YES NO Process system /
Well system
75 2003 Platform 47 YES 0.1 G Pig trap Significant leak NO NO Process system
77 2003 Platform 69 NO 1.2 G Standard flange Significant leak NO NO Process system
79 2003 Platform 56 YES 0.1 G Valve Significant leak YES NO Process system /
Open drain
80 2004 Platform 70 NO 0.16 L Filter Significant leak YES NO Process system
81 2004 Platform 71 NO 0.2 L Storage tank Significant leak NO NO Process system /
Storage
82 2004 Platform 64 YES 3 L Filter Significant leak YES NO Process system
83 2004 Platform 17 YES 2.8 G Standard flange Significant leak YES NO Flare system
84 2004 Platform 62 YES 17.2 L Steel pipe Significant leak YES NO Process system
85 2004 Platform 57 YES 0.71 G Hose Marginal leak YES NO Process system /
Well system
86 2004 Platform 69 NO 0.3 G Steel pipe Significant leak NO NO Process system
89 2004 Platform 43 YES 1.65 L Steel pipe Significant leak NO NO Process system
90 2004 Platform 46 YES 0.22 G Standard flange Significant leak NO NO Process system
91 2004 Platform 72 NO 0.1 L Valve Significant leak YES NO Process system

dataset 2015 01.01.2001 31.12.2014
[kg/s]
94 2004 Platform 22 YES 0.2 G Process vessel Significant leak YES NO Process system
96 2004 Platform 19 YES 10 G Standard flange Significant leak YES NO Flare system
Well system
101 2005 Platform 45 YES 0.1 G Valve Significant leak NO NO Process system /
Well system
104 2005 Platform 5 YES 240 L Standard flange Significant leak NO NO Process system
105 2005 Platform 8 YES 0.8 G Standard flange Significant leak YES NO Closed drain /
Process system
109 2005 Platform 57 YES 0.68 G Compressor Significant leak YES NO Process system
110 2005 Platform 2 YES 2 G Vent Significant leak YES NO Produced water /
Sea water / Open drain
113 2005 Platform 27 YES 8.03 G Hose Marginal leak YES NO Process system
114 2005 Platform 70 NO 1.7 L Producing well Marginal leak YES NO Well system

dataset 2015 01.01.2001 31.12.2014
[kg/s]
Flare system
118 2005 Platform 46 YES 0.5 G Valve Significant leak NO NO Process system
119 2006 Platform 8 YES 930 G Steel pipe Significant leak YES NO Flare system
122 2006 Platform 33 YES 0.15 G Hose Significant leak NO NO Process system
123 2006 Platform 62 YES 11.11 G Standard flange Marginal leak YES NO Closed drain
128 2006 Platform 39 YES 0.5 L Process vessel Significant leak YES NO Flare system
129 2006 Platform 44 YES 0.7 G Instrument Significant leak NO NO Process system
131 2006 Platform 61 YES 0.1 L Hose Significant leak YES NO Process system
132 2006 Platform 18 YES 80 G Standard flange Significant leak YES NO Flare system
138 2007 Platform 12 YES 0.15 G Steel pipe Significant leak YES NO Fuel gas system

dataset 2015 01.01.2001 31.12.2014
[kg/s]
142 2007 Platform 43 YES 2.5 L Valve Significant leak NO NO Process system /
Well system
143 2007 Platform 43 YES 1 G Valve Significant leak NO NO Process system /
Well system
144 2007 Platform 47 YES 1.93 G Standard flange Significant leak NO NO Fuel gas system
146 2008 Platform 55 YES 10 L Steel pipe Significant leak YES NO Process system /
Storage
150 2008 Platform 60 YES 0.2 G Steel pipe Significant leak YES NO Process system /
Well system
Flare system
152 2008 Platform 37 YES 0.26 G Steel pipe Significant leak NO NO Process system /
Flare system
153 2008 Platform 37 YES 0.5 G Standard flange Significant leak NO NO Process system /
Produced water?
155 2008 Platform 4 YES 2.8 L Valve Significant leak YES NO Process system /
Utility system
156 2008 Platform 22 YES 0.24 G Standard flange Significant leak YES NO Fuel gas system

dataset 2015 01.01.2001 31.12.2014
[kg/s]
159 2008 Platform 2 YES 0.8 G Steel pipe Marginal leak YES NO Process system /
Seal oil system
161 2009 Platform 5 YES 9 G Valve Significant leak NO NO Process system
165 2009 Platform 22 YES 0.5 G Valve Significant leak YES NO Fuel gas system /
Diesel system
166 2009 Platform 60 YES 6.84 G Instrument Significant leak YES NO Process system /
Gas lift system
167 2009 Platform 10 YES 0.815 L Instrument Significant leak YES NO Process system
Closed drain
170 2009 Platform 2 YES 0.45 L Valve Marginal leak YES NO Process system
Well system
172 2009 Platform 18 YES 1.5 G Valve Significant leak YES NO Flare system
173 2009 Platform 72 NO 2 L Steel pipe Significant leak YES NO Process system
176 2010 Platform 63 YES 0.276 G Valve Marginal leak YES NO Process system /
Well system

dataset 2015 01.01.2001 31.12.2014
[kg/s]
178 2010 Platform 69 NO 0.8 G Valve Significant leak NO NO Process system
Well system
184 2010 Platform 31 YES 0.5 L Standard flange Significant leak NO NO Process system
186 2010 Platform 63 YES 0.1 G Steel pipe Marginal leak YES NO Process system /
Well system
189 2010 Platform 72 NO 0.8 L Valve Significant leak YES NO Process system
190 2010 Platform 5 YES 0.1 G Pump Significant leak NO NO Process system /
Well system
197 2011 Platform 61 YES 0.58 L Steel pipe Significant leak YES NO Process system

dataset 2015 01.01.2001 31.12.2014
[kg/s]
201 2011 Platform 16 YES 0.34 G Valve Marginal leak YES NO Fuel gas system
202 2012 Platform 27 YES 16.9 G Standard flange Significant leak YES NO Flare system
204 2012 Platform 7 YES 0.17 G Compressor Significant leak YES NO Process system
205 2012 Platform 51 YES 0.48 L Valve Marginal leak YES NO Process system
206 2012 Platform 25 YES 230 L Valve Significant leak YES NO Process system
207 2012 Platform 73 NO 0.16 G Standard flange Significant leak YES NO Process system
Well system
209 2013 Platform 37 YES 0.39 G Valve Significant leak NO NO Process system /
Well system
213 2013 Platform 23 YES 20 G Compressor Significant leak YES NO Process system
214 2013 Platform 32 YES 0.9 G Valve Marginal leak NO NO Process system /
Well system
216 2013 Platform 55 YES 0.131 L Pump Significant leak YES NO Process system
217 2014 Platform 3 YES 0.15 G Steel pipe Significant leak YES NO Process system /
Flare system

dataset 2015 01.01.2001 31.12.2014
[kg/s]
220 2014 Platform 57 YES 20.8 L Vent Significant leak YES NO Closed drain /
Open drain system

Appendix B
Email used as reference
Report no: 105586/TN-2 Rev: Final A

1 Introduction ...................................................................................................................................B1
2 Email used as reference .................................................................................................................B1
Report no: 105586/TN-2 Rev: Final A Page Bi

1 Introduction
This appendix gives email used as reference in TN-2. The email is written in Norwegian.
2 Email used as reference

Hei,
Midlertidig bruk av slanger
Vi har hatt ei internmte med OMT- ekspertene i Safetec vedrrende modell for midlertidig bruk
av slanger;
Vi fokuserte i mtet p utfordringen knyttet til estimere lekkasjefrekvens for midlertidig bruk av
slanger ut fra den tilgjengelige dataen fra 3 Statoil innretninger. De tre installasjonene er ikke
ndvendigvis gode representanter for gjennomsnittet for Statoils innretninger p norsk sokkel. De
tre kan for eksempel alle ha et veldig hyt aktivitetsniv. Vi brukte derfor tilgjengelig informasjon
om hvordan deres aktivitetsniv forholder seg til gjennomsnittlig aktivitetsniv til lage et vektet
gjennomsnitt.
Vektene ble bestemt ved regne ut de tre installasjonenes relative aktivitetsniv for arbeidsordre
p normalt trykksatt utstyr, B1-B4, i forhold til gjennomsnittet for alle Statoils innretninger p
norsk sokkel. Alts gjorde vi en implisitt antakelse om at for den enkelte installasjon vil
aktivitetsniv for B6 forholde seg til gjennomsnittlig niv p samme mte som B1-B4. Dette vet vi
ikke er riktig for den enkelte installasjon, fordi B6-niv avhenger sterkt av design, men vi fant det
allikevel rimelig tro at denne vektingen gir et bedre estimat enn ved ikke vekte snittet.
1) Vektet snitt B6 operasjoner i ret for de 3 installasjonene :(260+371+12)/3=214

2) Lekkasjefrekvens B6 Statoil per r fra tabell 4.2: 1.07E-02
3) Lekkasjefrekvens per slangeoperasjon flger da med 1.07E-02/214=5.0E-05
Dette er det beste vi kan gi som input til ein aktivitetsbasert modell for bruk av midlertidige
slanger. Vi mener dette er en ok fremgangsmte, men ser at det er usikkerhet her. Vi kan
utdypetallmaterialet bak det vektet snittet dersom dette er av interesse.
QA av norsk lekkasjedata.
Lekkasje-ID 20 og 22 som vi har latt st som ukjent p type utstyr er klassifisert som B6 lekkasjer i
BORA. B6 er definert som ; Maloperation of temporary hoses. Vi har ikke granskingsrapport p
disse lekkasjene, s vi kan dessverre ikke g tilbake verifisere at dette virkelig er B6 lekkasjer.
Men det er alts gjort ein vurdering ein gang der dei har landa p at dette er B6 lekkasjar.
Berra ta kontakt om det er spm/kommentarar til dette.
God helg!
Mvh
Jon Andreas Hestad Office: +47 415 14 647 (Bergen)

Senior Safety Engineer Direct: +47 55 55 10 90
Jon.Andreas.Hestad@safetec.no
www.safetec.no | www.abs-group.com
Report no: 105586/TN-2 Rev: Final A Page B1

Working together
for a safer world
UKCS-data
TN-3
Technical note for:

Technical note no: 105586/TN-3 Rev: Final B

Date: 18 March 2016
1 Introduction .....................................................................................................................................1
2 Recorded incidents at UKCS relevant for the modelled leak scenarios ..............................................1
2.1 Extracting relevant process leaks fed through process system ................................................1
2.1.1 Filters used to extract data .................................................................................................. 1
2.1.2 Extracted data for process leaks fed through process systems .............................................. 5
2.2 Extracting relevant process leaks fed through utility system .................................................11
2.2.1 Description of filters ......................................................................................................... 12
2.2.2 Extracted data for process leaks fed through utility systems ............................................... 12
2.3 Extracting relevant process leaks from well system ..............................................................14
2.3.1 Gas leaks from oil wells .................................................................................................... 14
2.3.2 Oil leaks from oil wells ...................................................................................................... 14
2.3.3 Leaks from gas wells ......................................................................................................... 14
2.3.4 Leaks from X-mas tree ...................................................................................................... 15
2.3.5 Extracted data for leaks from well system.......................................................................... 15
2.4 Summary of relevant leaks extracted from HCRD .................................................................17
3 Exposure database .........................................................................................................................20
3.1 Process equipment ..............................................................................................................20
3.2 Well head ............................................................................................................................21
4 Calculation of leak frequencies based on HCR-data and trends in data material ............................23
5 Complementary cumulative hole size distributions and leak rate distributions based on HCRD ......26
6 Uncertainty and quality of HCR-data ..............................................................................................30
6.1 Incident data .......................................................................................................................30
6.2 Exposure data......................................................................................................................32
6.3 Concluding remark ..............................................................................................................32
7 References .....................................................................................................................................33
Appendix A HCR databasis

Appendix B Extracted data from HCR databasis
Appendix C Complementary cumulative hole size distributions and leak rate distributions based
on HCRD

Page ii
1 Introduction
This technical note describes incident data and population data extracted from HCRD (Chapter 2
and 3) and how these data is used to estimate leak frequencies and hole size distributions for the
equipment types covered by the model. Together with incident data and population data (ex-
posure data) from the Norwegian Continental Shelf (NCS), these data are used as basis for para-
metrization of the leak frequency model for the defined leak scenarios described in TN-4. How
the data is used to parametrize the mathematical functions described in TN-5 is described in TN-6.
Information about offshore releases of hydrocarbons at United Kingdom Continental Shelf
(UKCS), are collected in Hydrocarbon Release Database (HCRD). The database is operated by
Health and Safety Executive (HSE).
Lilleaker Consulting AS (hereafter denoted Lilleaker) has built a databasis in excel format with all
HCR-data, and developed additional data fields (based on the existing data fields), filters and
tools for data analysis. The HCR databasis is documented in Appendix A, which contains
Lilleakers documentation of the HCR-data, documentation of the developed databasis and also
general assessments of the data fields in HCRD.
The developed databasis has been made available to all project participants, but is not publicly
available. Important parts of the data extracted from HCR data is given in Appendix B. Note also
that all data in the databasis, except exact hole sizes for holes >100 mm and exact equipment
dimensions are publically available as described in Appendix A. Exact hole sizes for holes >100
mm and exact equipment dimensions have been made available to this project by HSE.
expressions.
2 Recorded incidents at UKCS relevant for the modelled

leak scenarios
In total 4561 events occurring in the period Q3 1992 - Q1 2015 are recorded in HCRD. Not all of
the incidents are relevant for the defined leak scenarios (see TN-4). A thorough analysis has been
necessary to extract the relevant incidents for the model. In this chapter, filters are defined and
described to explain how the relevant incidents are filtered out. This is done separately for
process leaks fed through process systems, process leaks fed through utility systems, producing
well leaks and gas lift well leaks, in Chapter 2.1, 2.2 and 2.3, respectively.
2.1 Extracting relevant process leaks fed through process system

This chapter describes the applied filters to extract process leaks fed through process systems
from HCRD. Further the number of incidents extracted by applying the filters is given in detail in
Appendix B, while a summary of the extracted data is presented here.
2.1.1 Filters used to extract data
An illustration of the applied filters is given in Figure 2.1. The figure shows the number of
incidents removed from the databasis in each filter operation, and how many that remains in
each step. The resulting databasis contains 2855 recorded incidents from the period Q3 1992 -
Q1 2015, and 1597 recorded incidents from the period Q1 2001 - Q1 2015. These incidents are
further divided into the following categories:
Incidents with total recorded released amount 10 kg and >10 kg.
Incidents with recorded initial pressure 0.01 barg and >0.01 barg.

Page 1
Incidents with recorded hole size 1 mm, >1 mm, and incidents where the recorded hole
size is recorded as N/A (Not Applicable). In Appendix A in Lilleakers report, given in TN-3
Appendix A, the HCR definitions of the data fields are presented. For hole diameters it is
stated: It is important to note that N/A in this field indicates that hole size is not applicable
to the mode of release involved. An example of incidents from HCRD where hole size is
recorded as N/A is if oil is carried up the HP flare, where not all of the oil is burned and some
drops as droplets to the sea or platform topside.
The detailed results are given in Appendix B.
The filters applied to HCRD to extract relevant process leaks fed through process systems are des-
cribed in detail in the following sections. An overview of the evaluated data-fields is given in
Table 2.1.
Figure 2.1 - Illustration of the filters used to extract relevant process leaks fed through process
systems from HCRD. The numbers with green font represent incidents that are kept after the filter
is applied. The numbers in red font are the number of incidents that are taken out. The number at
the left side of the slash are resulting from the period Q3 1992 - Q1 2015, while the number at the
right side of the slash are resulting from the period Q1 2001 Q1 2015.

Page 2
Table 2.1 - HCR data fields evaluated and described in more detail in the below sub-sections.
HCR HCR data field HCR Description

field no.
2 CATEGORY Installation type: FIXED, MOBILE, SUBSEA. The installation
may have a subsea satellite (recorded in field 16 subsea)
19 PROCESS This is the type of Hydrocarbon released, i.e. NON-PROCESS,
OIL, CONDENSATE, GAS and 2-PHASE
28 SYSTEM This field contains either a full description of the system
involved or a Drilling or Well Operation activity description
where appropriate.
32 EQUIPMENT This gives the full equipment item description. For
Drilling/Well Operations activities (see item 28 above) this
will be left blank.
43 HAZ_CLASS This field contains the Hazardous Area Classification for the
location of the incident, where 1 and 2 represent areas 1
and 2 respectively, and 3 represents unclassified.
47 MOD_VOLUME This contains the volume of the module involved, in m3, and
will show NOT KNOWN where not reported.
53 INVENTORY This is the isolatable hydrocarbon inventory contained in the
system, in kg. And will show NOT KNOWN where not
reported.
58 DETECTION_OTHER Leak detected by other means
2.1.1.1 Relevant installations (CATEGORY)

HCRD distinguish on 3 different types of installations: Fixed, mobile and subsea installations. Only
incidents at fixed installations are regarded as relevant for the model.
2.1.1.2 Relevant leaks medium (PROCESS)
HCRD distinguish on NON-PROCESS, OIL, CONDENSATE, GAS and 2-PHASE leaks. Leak medium
categorized as non-process is regarded as not relevant for the model.
2.1.1.3 Relevant systems (SYSTEM)
HCRD describes the system involved. The systems regarded as relevant and irrelevant for the
model are listed in the table below.

Page 3
Table 2.2 - Systems (as defined in HCRD) regarded as relevant and not relevant for the process
leaks fed through process system. For definitions of the systems it is referred to Appendix A
Relevant systems Not relevant systems
Export Blowdown and flare
Metering Subsea well
Flowlines Vent
Compression Closed drain
Fuel gas Open drain
Processing Surface well
Import Well control
Separation Turbines
Drilling
Utilities
2.1.1.4 Relevant equipment (EQUIPMENT)

The equipment that is regarded as relevant and irrelevant, for the model is listed in Table 2.3.
Note that the naming convention is in accordance with HCRD.
Table 2.3 - Equipment regarded as relevant and not relevant for the model. Valves, flanges and
pipes are given in HCRD as three equipment size intervals; small (3), medium (3-11) and large
(>11). The model equipment naming is given in parenthesis
Table Heading
Actuated valve L (Valve)
Actuated valve M (Valve)
Actuated valve S (Valve)
Manual valve L (Valve)
Manual valve M (Valve) Degasser
Manual valve S (Valve) Expanders
Centrifugal Compressors (Centrifugal Compressor) Drain
Reciprocating Compressor (Reciprocating Compressor) Flexible pipelines
Filters (Filter) Pipeline valve
Flanged joints L (Standard flange) Flexible risers
Flanged joints M (Standard flange) Steel risers
Flanged joints S (Standard flange) Steel pipeline
Heat exchanger plate (Plate heat exchanger) Turbines
Heat exchanger HC in tube (Tube side heat exchanger) Xmas trees
Heat exchanger HC in shell (Shell side heat exchanger) BOP
Fin fan cooler (Air-Cooled Heat Exchanger) Shale shakers
Instruments (Instrument) Recompressor
Pig traps (Pig trap) Wellhead
Process vessel (Process vessel) Mud pumps
Centrifugal pump (Centrifugal pump) Mud tanks
Reciprocating pump (Reciprocating pump) Workover

Page 4
Table Heading
Steel piping large (Steel pipe) #N/A
Steel piping medium (Steel pipe)
Steel piping small (Steel pipe)
Atmospheric vessel (Atmospheric vessel)
Flexible piping (Flexible pipe)
2.1.1.5 Relevant area classification (HAZ_CLASS)

No incidents are removed from the database based on recorded area classification.
2.1.1.6 Relevant module volume
The term module is not defined in HCRD, but it is stated: 3000m3 explosive clouds are enough
to fill an entire module or deck area. Module volumes are sometimes reported to be very small,
maybe inside confinements such as separate rooms (e.g. for pumps) or under hood of turbines.
No incidents are removed from the database based on recorded module volume.
2.1.1.7 Relevant inventory
Many recorded inventories are reported being very small. One could claim that the inventory of a
standard isolatable segment should be significant in order to the leak to be relevant for the
model. However, incidents are not removed from the database based on recorded inventory.
2.1.1.8 Relevant detection method
The recorded detection method may indicate that the leak was not a process leak. For instance;
ROV detection or pressure drop may indicate subsea leak, which is possibly indicate leaks that
should be considered irrelevant for the model. However, incidents are not removed from the
database based on recorded detection method.
2.1.1.9 Hole size
The existing model is valid for hole sizes >1mm. The uncertainty related to hole sizes <1mm is
significant, and the same model validity range as assumed in the previous model is suggested for
the updated model. However, these incidents are included in the analysis, but separated from
incidents with hole size > 1 mm.
2.1.1.10 Initial leak rate
Incidents are not removed from the database based on initial leak rate boundary.
2.1.2 Extracted data for process leaks fed through process systems
The data extracted from HCRD by applying the filters described in Section 2.1.1 Figure 1, are
given in detail in Appendix B. Figures that show the most important observations related to
process leaks fed through process systems are given in the below figures.
In total there are 2855 relevant incidents in the period Q3 1992 Q1 2015 in HCRD, and 1597
relevant incidents in the period Q1 2001 Q1 2015. About 50 % of these incidents are recorded
with hole size 1 mm. Also a significant fraction of the leaks are recorded with a total leaked
quantity 10 kg, which are classified as Marginal leaks in accordance with the definitions in TN-
4. Figure 2.2 shows the number of relevant Marginal and Significant leaks with hole size > 1 mm
or N/A and hole size 1 mm for the periods Q3 1992 Q1 2015 and Q1 2001 Q1 2015. Note
that Significant leaks with initial pressure >0.01 barg and 0.01 barg are given separately and
shows that the number of significant leaks with initial pressure 0.01 barg is low. This is also
seen in Figure 2.3 that shows the relative contribution from all these leak scenarios. Figure 2.4
gives the fractions of relevant leaks recorded in HCRD with hole size >1 mm or with hole size
N/A, for Marginal and Significant leaks.

Page 5
Figure 2.5 and Figure 2.6 give the equipment type distribution for the period Q3 1992 Q1 2015
and Q1 2001 Q1 2015 for Significant and Marginal leaks, respectively, while Figure 2.7 gives
the equipment type distribution for Marginal and Significant leaks for the period Q3 1992 Q1
2015.
Reported leaks at NCS (see TN-2) only comprise leaks with initial leak rate >0.1 kg/s. Therefore it
is of interest to see the fraction of incidents recorded at UKCS that has an estimated initial leak
rate >0.1 kg/s. This is given in Figure 2.8 for significant leaks with hole size > 1 mm recorded in
the period Q3 1992 Q1 2015.
Figure 2.2 - Number of process leaks fed through process systems recorded on UKCS relevant for
the defined leak scenarios

Page 6
Figure 2.3 - The fraction of leaks that are relevant for the defined leak scenarios
Figure 2.4 - Fractions of relevant leaks recorded in HCRD with hole size >1 mm or with hole size
N/A, for Marginal and Significant leaks

Page 7
Figure 2.5 - Equipment type distribution for Significant leaks, given both for the time period Q3
1992 Q1 2015 and Q1 2001-2015
Figure 2.6 - Equipment type distribution for Marginal leaks, given both for the time period Q3
1992 Q1 2015 and Q1 2001- Q1 2015

Page 8
Figure 2.7 - Equipment type distribution for Significant and Marginal leaks for the period Q3 1992
Q1 2015
Figure 2.8 - Fraction of Significant leaks in the period Q3 1992 Q1 2015 that has initial leak rate
0.1 kg/s, and > 0.1 kg/s. Only hole sizes >1 mm or N/A are included

Page 9
2.1.2.1 Effect of reducing the data collection period
HCR-data is available from Q3 1992, but the latest data are most likely more representative for
the future than the oldest data. Therefore it is of interest to study the number of incidents
remaining if the start date for the collection period is changed. This is given in Figure 2.9 for all
steps in the defined filter in Figure 2.1, and also if incidents with recorded hole size < 1 mm are
removed. The numbers of incidents are reduced linearly, indicating that the number of leaks per
year is relatively constant before 2001. This is confirmed in Figure 2.11 that gives the number of
relevant recorded leaks in HCRD in for every year in the period 1993 2014. The years 1992 and
2015 are not included as data for the full year is not available. The figure displays a decreasing
trend after 2004. As corresponding exposure data are not given per year, leak frequency trend
with time cannot be analysed.
For every step in the defined filter in Figure 2.1, Figure 2.10 gives the fraction of process incidents
as a function of the first year in the data collection period relative to using 1992 as the first year.
All filter steps show similar trend (they are on top of each other) except for the hole size filter,
indicating that the frequency of process leaks at fixed installations, from relevant systems and
from relevant equipment is constant in before 2001. The figure also shows that the fraction of
these leaks with hole size >1 mm is decreasing, which indicates that there is a decreasing trend in
frequency for leaks relevant for modelling of process leaks in Quantitative Risk Analysis (i.e. initial
leak rate > 0.1 kg/s).
Figure 2.9 - Number of process leak incidents left after the applied filters as a function of the first
year in the period of collected data (end year of period is 2015)

Page 10
Figure 2.10 - Fraction of process incidents left after the applied filters as a function of the first year
in the period of collected data (end year of period is 2015)
Figure 2.11 - Number of relevant process incidents recorded in the period 1993-2014. The total
number of recorded leaks in this period is 2826
2.2 Extracting relevant process leaks fed through utility system

In this chapter, the filters used to extract process leaks fed through utility systems are described.
The number of incidents extracted by applying the filters is given in detail in Appendix B, while a
summary of the extracted data is presented here.

Page 11
2.2.1 Description of filters
In the following sub-sections the filters used to extract process leaks fed through vent, drain and
flare are described. These scenarios are in accordance with the leak scenarios covered by the
model as described in TN-4. Note that process leaks fed through injection systems should also be
included. In HCRD there is one incident that could be a relevant process leak fed through an
injection system. However, this incident has been disregarded. It is unclear whether this leak is
relevant. In any case, the contribution from this single incident is negligible.
2.2.1.1 Vent leaks
To extract incidents where process fluid has been released through vents, due to overfilling or
other maloperations that represent a potential major accident hazard have been done by
applying the following filter to the HCR-data
Process: All except non-process
Category: Only fixed installation
My system: Only vent
My equipment: All relevant equipment in Table 2.3
Act pressure/max_pressure: Only 1-10. This represents leaks where the recorded pressure is
higher than the design pressure, which indicates that the incident occurred due to
maloperation.
2.2.1.2 Drain leaks
To extract incidents where process fluid has been released through drain systems, the following
filter to the HCR-data
My system: Closed drain + Open drain
My equipment: All relevant equipment in Table 2.3.
2.2.1.3 Flare leaks
To extract incidents where process fluid has been released through flare systems, the following
filter to the HCR-data
My system: Blowdown & Flare
My equipment: All relevant equipment in Table 2.3.
2.2.2 Extracted data for process leaks fed through utility systems
The data extracted from HCRD by applying the filters described in Section 2.2.1, are given in de-
tail in Appendix B. A summary is given in the following figures. In total 253 leaks with hole size >
1 mm (or N/A) are included for the period 1992-2015, while for the period 2001-2015, the corre-
sponding number is 145 leaks. The distribution per leak scenario is shown in Figure 2.12 and
Figure 2.13.

Page 12
Figure 2.12 - Number of process leaks fed through utility systems recorded on UKCS considered
relevant for the defined leak scenarios
Figure 2.13 - Distribution of process leaks fed through utility systems

Page 13
2.3 Extracting relevant process leaks from well system
In this section the filters used to extract gas lift well leaks and producing well leaks (see TN-4 for
definition of gas lift well leak and producing well leak) from HCRD are defined. Filtering of
relevant incidents is done by extracting
gas leaks from oil wells
oil leaks from oil wells
leaks from gas wells
leaks from X-mas trees
separately by the filters described in the below sub-chapters. Gas leaks stemming from oil wells
are assumed to be leaks from the gas lift system, while all other leaks are assumed to be leaks
from the producing well. Note that the incidents extracted has not been studied in detail, and
therefore it is a significant uncertainty related to the estimation of well leak frequencies based on
the extracted incidents from HCRD.
2.3.1 Gas leaks from oil wells
Gas leaks from oil wells are assumed to be leaks from the gas lift system of the well. To extract
these incidents from HCRD, the following filter is applied:
Process: Only gas
My system: Only Surface oil well
My equipment: Only Wellhead
Operational mode: All except well services (see definition of OP_MODE in Appendix A).
2.3.2 Oil leaks from oil wells
Oil leaks from oil wells are assumed to be leaks from the producing well. To extract these
incidents from HCRD, the following filter is applied:
Process: All except non-process and gas
My system: Only Surface oil well
Operational mode: All except well services (see definition of OP_MODE in Appendix A).
2.3.3 Leaks from gas wells
Leaks from gas wells are assumed to be leaks from the producing well. To extract these incidents
from HCRD, the following filter is applied:
My system: The following systems are included
o Surface gas injection well
o Surface gas producing well
o Surface well other
Operational mode: All except well services, welloptree and drillgas (see definition of
OP_MODE in Appendix A).

Page 14
2.3.4 Leaks from X-mas tree
Both oil and gas leaks from X-mas tree are assumed to be leaks from the producing well. To
extract these incidents from HCRD, the following filter is applied:
My system: The following systems are included
o Surface gas injection well
o Surface gas producing well
o Surface oil well
o Surface well other
My equipment: X-mas tree
Operational mode: All except well services, welloptree and drillgas (see definition of
OP_MODE in Appendix A).
2.3.5 Extracted data for leaks from well system
The data extracted from HCRD by applying the filters described in Section 2.2.1 - 2.3.4, are given
in detail in Appendix B. In total 100 (17 gas lift leaks and 83 producing well leaks) incidents are
extracted for the period Q3 1992 Q1 2015 while 38 (9 gas lift leaks and 29 producing well
leaks) incidents from the period Q1 2001- Q1 2015 are identified as relevant for gas lift well leaks
and producing well leaks. In Figure 2.14 and Figure 2.15, the leaks are sorted with respect to the
defined Marginal and significant leak scenario and grouped depending on hole size. .
The results show that there has been a considerable decrease in leaks originating from well in the
period after 2001. It should also be noted that the fraction of Marginal leaks is larger than for
process leaks. The relative reduction in leaks after 2001 is most prominent for significant leaks,
which results in a high fraction of Marginal leaks for the period after 2001. Moreover, the
fraction of leaks resulting from a hole having a diameter less than 1 mm is larger than for leaks
from process systems. It has not been attempted to explain the causes for this observation, i.e.
the difference in fraction Marginal and Significant leaks originating from wells.

Page 15
Figure 2.14 - Extracted leaks from well system with hole size >1 mm or N/A
Figure 2.15 - Extracted leaks from well system with hole size 1 mm

Page 16
2.4 Summary of relevant leaks extracted from HCRD
This chapter gives a summary of the extracted incidents for process leaks fed through process
system and utility system (Vent, drain and flare) and leaks from well systems. The detailed
number of recorded leaks, as well as the exposure data is given in Appendix B. The total number
of recorded process leaks and leaks from well system extracted from HCRD is given in Figure
2.16, while the fraction of leaks fed through process system, vent, drain and flare system and
well system is given in Figure 2.17. Figure 2.18 and Figure 2.19 give the equipment type
distribution for Significant and Marginal leaks for the period Q3 1992 Q1 2015 and Q1 2001
Q1 2015. All process leaks and leaks from wells are included. Figure 2.20 and Figure 2.21 give
also the equipment type distribution for the period Q3 1992 Q1 2015 and Q1 2001 Q1 2015,
but the figures also include the equipment size distribution where only incidents recorded with
initial leak rate >0.1 kg/s are included. This corresponds to how leaks are logged on NCS (see
TN-2).
Figure 2.16 - Total number of extracted process leaks from HCRD. The leaks are categorized into
Marginal and Significant leaks. Only hole sizes > 1mm (or N/A) are included

Page 17
Figure 2.17 - Relative contribution from the same scenarios and incidents as included in Figure 2.16
Q1 2015. All process leaks and leaks from well system are included. Only hole sizes >1 mm (orN/A)
are included

Page 18
Q1 2015. All process leaks and leaks from well system are included. Only hole sizes >1 mm (orN/A)
are included
Figure 2.20 - Equipment type distribution for the period Q3 1992 Q1 2015. All process leaks and
leaks from well system are included. The blue columns corresponds to the blue columns in Figure
2.18, while the red columns only includes incidents with initial leak rate >0.1 kg/s. This corresponds
to the leaks logged on NCS. Only hole sizes >1 mm (or N/A) are included

Page 19
Figure 2.21 - Equipment type distribution for the period Q1 2001 Q1 2015. All process leaks and
leaks from well system are included. The blue columns corresponds to the blue columns in Figure
2.19, while the red columns only includes incidents with initial leak rate >0.1 kg/s. This corresponds
to the leaks logged on NCS. Only hole sizes >1 mm (or N/A) are included
3 Exposure database
Chapter 3.1 and 3.2 presents the population data extracted from HCRD for process equipment
and wellheads, respectively. Chapter 3.3 presents known issues generating uncertainty related to
the exposure data in HCRD.
3.1 Process equipment

The exposure data (population data) for relevant process equipment types extracted from HCRD
is given in detail in Appendix B. Figure 3.1 gives the exposure data for relevant equipment types.
Note that the scale of the y-axis is logarithmic. The population data are used for estimating leak
frequencies per equipment per year as described in Chapter 0.
HCRD defines one flange face as one flange. In the suggested counting guideline (TN-5 Appendix
A) which is in accordance with population data extracted from QRAs for installations on the NCS,
two flange faces are counted as one flanged joint. In order to adjust for this difference, the ex-
posure data extracted from HCRD for flanges is divided by a factor 2. This is not entirely correct
as some flanges consist of only one flange face (e.g. blinded flanges for temporary mounting of
equipment). The number of flange years at UKCS will therefore be slightly underestimated using
a factor of 2.

Page 20
Figure 3.1 - Exposure data for relevant equipment types. For steel pipe and flexible pipe the
exposure data is given as the number of equipment year meters. Note that the y-axis has
logarithmic scale. Exact values are given in Appendix B
3.2 Well head

The number of wellhead years extracted from HCRD is given in detail in Appendix B for:
Gas injection wellhead
Gas producing wellhead
Oil producing wellhead
Other wellhead.
This is also presented in Figure 2.4. The figure shows the exposure data both for the period Q3
1992- Q1 2001 and the period Q1 2001 Q1 2015. Table 3.1 and Figure 2.5, gives the
estimated exposure data for gas lift well and producing well. The following assumptions are
made:
1. All types of wellheads given above are relevant for producing wells
2. The number of gas lifted wells on UKCS is not available. However, an estimate is established
based on the SINTEF offshore blowout database, Ref. /1/. For US GOM OCS, the percentage
of gas lifted wells is from 20 % to 63 % in the period 1992-2012. 50 % is suggested for
UCKCS for the period 2001-2015.
Table 3.1 - Exposure data for well heads extracted from HCRD
Exposure data
Well head type 1992-2015 2001-2015
Gas lift well 5953 3515
Producing well 28081 17670

Page 21
Figure 3.2 - Exposure data for relevant types of wellhead. Note that the y-axis has logarithmic scale
Figure 3.3 - Estimated exposure data for gas lift wells and producing wells

Page 22
4 Calculation of leak frequencies based on HCR-data and
trends in data material
Based on the number of leak incidents for equipment type , denoted , and the number of
equipment years (exposure data) for equipment type , denoted the leak frequency is
calculated as
(1)
=

The estimated leak frequency per component based on HCR-data are presented in detail in App-
endix B. Figure 4.1 and Figure 4.2 give the estimated process leak frequency for hole size >1 mm
(or N/A) for Marginal and Significant leaks, respectively. Figure 4.4 shows the ratio obtained
when the total leak frequency for all hole sizes is divided by the leak frequency for hole size >1
mm for Significant leaks Figure 4.3 shows the same result for Marginal leaks. This ratio is
denoted K1mm in TN-6 when the model is parameterized based on the HCR data.
Figure 4.1 - Estimated process leak frequency for Marginal leaks with hole size >1 mm or N/A

Page 23
Figure 4.2 - Estimated process leak frequency for Significant leaks with hole size >1 mm or N/A
Figure 4.3 - Marginal leaks; Total leak frequency divided by leak frequency for hole size >1 mm

Page 24
Figure 4.4 - Significant leaks: Total leak frequency divided by leak frequency for hole size >1 mm
(ratio denoted K in TN-6)
1mm
Figure 4.5 - Fraction of the total leak frequency distributed on Marginal leaks and Significant leaks.
For significant leaks the contribution in terms of system pressure when leak occurs is presented
(above and below 0.01 barg)

Page 25
5 Complementary cumulative hole size distributions and
leak rate distributions based on HCRD
In order to establish hole size distributions based on HCRD, relevant incidents has to be extracted.
The filter defined in Figure 2.1, is used as basis, but hole sizes < 1mm or hole sizes recorded as
N/A are not included. The hole size distributions will first and foremost be used to estimate the
frequency for holes resulting in Significant leaks. Hence, incidents recorded with total released
quantity <10 kg and leaks with initial pressure < 0.01 barg are not included in Filter 1, (see Table
5.1 below). This is considered to be the most relevant filter for parameterization of the hole size
distributions in the model.
In order to investigate the effect of including other leaks, i.e.
process leaks fed through utility systems;
leaks recorded with total released quantity <10 (Marginal leaks) and
leaks with initial pressure <0.01 barg
alternative filters denoted filter 2 and filter 3 are established. Filters extracting incidents from the
period Q3 1992- Q1 2015 are denoted a, while filters from the period Q1 2001- Q1 2015, are
denoted b. The number of incidents included as basis for the hole size distributions for these
two periods are given in Figure 5.1 and Figure 5.2, respectively. Filter 4 is defined to produce
initial leak rate distributions based on the same type of incidents as the initial leak rate
distributions based on NCS data (i.e. leaks are filtered based on initial leak rate, and not hole size)
are based on (see TN-2). The number of incidents included as basis for the initial leak rate
distributions for the two periods are given in Figure 5.3 and Figure 5.4. For simplicity the initial
leak rate distributions are denoted leak rate distributions.
All complementary cumulative hole size distributions and leak rate distributions based on HCRD
are given in Appendix C. An example of a hole size distribution is given in Figure 5.5, where all
equipment types are included. In general, filter 3 result in a larger fraction large holes compared
to filter 2. Filter 2 result in a larger fraction large holes compared to filter 1. However, for some
equipment types, the situation is the other way around, and for many equipment types the
difference between the filters is marginal.
The complementary cumulative hole size distribution for all equipment types based on recorded
hole sizes in HCRD is shown in Figure 5.6. Separate leak rate distributions are plotted for gas (gas
and 2-phase), liquid (oil and condensate) and gas and liquid altogether (denoted G, L and G&L,
respectively). Note that the initial leak rates are calculated by Lilleaker based on hole size and
available process conditions in HCRD (using the same formulas as in the validation model
presented in TN-6). There is however good reasons to question the quality of the data put as
basis for the calculations (see Chapter 6.1).
Table 5.1 - Filters used to extract incidents (hole sizes) as basis for recorded hole size distributions
based on HCRD
Filter Description
Filter 1 All relevant process leak incidents in the HCR-data as defined in Figure 2.1 and
well system leaks as defined in Chapter 2.3 are included, except:
Incidents recorded with pressure <0.01 barg
Incidents recorded with total released quantity <10 kg
Incidents recorded with hole size <= 1 mm
Incidents recorded with hole size N/A
This filter is put as basis for hole size distributions in the model development.

Page 26
Filter Description
Filter 2 All relevant process leak incidents in the HCR-data as defined in Figure 2.1,
relevant leaks fed through utility systems as defined in Chapter 2.2 and relevant
well releases as defined in Chapter 2.3 are included, except:
This filter is defined to analyse the effect of including process leaks fed through
utility systems and well systems as basis for hole size distributions.
relevant utility leaks as defined in Chapter 2.2 and relevant well releases as
defined in Chapter 2.3 are included, except:
This filter is defined to also analyse the effect of including incidents recorded with
pressure <0.01 barg, and incidents recorded with total released quantity <10 kg.
relevant utility leaks as defined in Chapter 2.2 and relevant well releases as
defined in Chapter 2.3 are included, except:
Incidents recorded with initial leak rate <0.1 kg/s
This filter is defined to establish leak rate distributions based on the same type of
incidents as the leak rate distributions based on NCS data are based on.
Figure 5.1 - The number of incidents included as basis for the recorded hole size distributions for
the period Q3 1992- Q1 2015. Filters extracting incidents from this period are denoted Filter 1a,
Filter 2a and Filter 3a. The filters are defined in Table 5.1

Page 27
the period Q1 2001- Q1 2015. Filters extracting incidents from this period are denoted Filter 1b,
Filter 2b and Filter 3b. The filters are defined in Table 5.1
the period Q3 1992- Q1 2015. Filters extracting incidents from this period are denoted Filter 1a,
Filter 2a and Filter 3a. The filters are defined in Table 5.1

Page 28
the period Q1 2001- Q1 2015. Filters extracting incidents from this period are denoted Filter 1b,
Filter 2b and Filter 3b. The filters are defined in Table 5.1
Figure 5.5 - Complementary cumulative hole size distribution for all equipment types, based on
recorded hole sizes in HCRD

Page 29
Figure 5.6 - Complementary cumulative leak rate distribution for all equipment types, estimated
based on hole sizes and process conditions recorded in HCRD. Separate curves are given for gas
(gas and 2-phase), liquid (oil and condensate) and gas & oil, denoted G, L and G&L, respectively
6 Uncertainty and quality of HCR-data

There are uncertainty related to the recorded hole sizes and recorded process conditions in
HCRD. There are also a known deficiencies related to the exposure data.
6.1 Incident data

The following understanding of the quality of the HCR-database was achieved in the project
meeting 04.09, Ref. /2/: The registration of incidents in HCRD is voluntary, but it is expected that
the general industry practice is that incidents are registered. Thus, it is reasonable to believe that
the database is quite complete in terms of number of incidents. In the initial phase of the project,
upgrading the database has discovered some inconsistencies in the raw data and the publicly
available HCR-data. This may be due to inadequate procedures for compiling the data. Some
issues identified are:
Data fields were not the same in the two data sets (raw data and the publicly available HCR-
data)
The data sets had two ways of assessing the hole size
o Calculated hydraulic hole size diameter.
o Measured hydraulic hole size diameter
Uncertainty which of the two data sets that contains the most correct value.
The hole size recorded in HCRD is of particular importance for the model development. The mo-
del are based on hole size distributions, and hole sizes are not recorded as part of the registered
leaks at installations on the NCS data. Hence HCRD is the only available data source where hole
sizes are available. In the HCR-definitions, the data field HOLE_DIAM, which gives the hole dia-
meter used as basis for the model, is defined as follows (see also Appendix A in Appendix A):

Page 30
HOLE_DIAM - This is the hydraulic equivalent hole size, deduced from d = 4A/p, in mm. Where
d is the diameter of the hydraulic equivalent hole, A is the cross-sectional area of the actual hole
in mm2, and p is the wetted perimeter of the actual hole in mm. It is important to note that N/A
in this field indicates that hole size is not applicable to the mode of release involved.
Note that the definition does not state whether the diameter is measured or calculated, but the
project meeting 04.09 indicates that some are calculated and some are measured. The methodo-
logy for calculating the hole sizes are not stated in HCRD. Lilleaker has calculated the initial leak
rate based on hole size and process conditions at the onset of the leak (using the same equations
as in the validation model presented in TN-6). In Figure 6.1, the ratio between calculated initial
leak rate and average leak rate is plotted for all relevant process leaks fed through process
systems (2855 incidents, see Figure 2.1). A similar figure is given in Appendix A for all leaks in
HCRD. Figure 6.1 shows that ratio for about 2/3 of all incidents is 1. This means that the
estimated initial leak rate is less than the average leak rate. This demonstrates that some data
fields are incorrect. These uncertainties must be accounted for when interpreting the data and
using the data to parameterize the leak frequency model.
Figure 6.1 - Ratio between calculated initial leak rate and average leak rate. The x-axis gives the
fraction of the total number of relevant process leaks fed through process systems (2855 incidents,
see Figure 2.1). A similar figure is given in Appendix A for all leaks in HCRD

Page 31
6.2 Exposure data
The quality of the HCRD exposure data is not complete, (Ref. /2/). The following aspects must be
considered when using the data for estimation of leak frequency per component:
a Several installations at UKCS are not included in the population database. It is not known
how many, but it is judged that the number of installations not included is few compared to
the total number included
b The population data has not been maintained and updated after 2006
c The procedures for update of population data is unclear in terms for responsibility for
maintenance of the data
d There is most likely inconsistency in the way equipment is counted on the various
installations (e.g. how instrument connections are counted with regard to flanges and valves
associated with instruments)
e Upgrade of the population data basis has been initiated by HSE and will be finalized in about
6 months.
6.3 Concluding remark

Based on above, the overall assessment of the HCR-database is that:
Leak frequencies per component based on HCRD will overestimate the underlying leak
frequency, and
The hole size distributions derived from HCRD is not completely representative for the
underlying hole size distributions. It is not possible to evaluate whether the actual underlying
hole size distribution is shifted towards smaller or bigger holes.
These aspects must be taken into account when HCR data is compared with NCS data, and when
the HCR data is applied for parameterization of the leak frequency model.

Page 32
7 References
offshore blowout database 2014, 17 March 2015, Report No: 19101001-8/2015/R3 Rev:
Final
/2/ Minutes of meeting from Project meeting 04.09.2015.

Page 33
Appendix A
HCR databasis
Report no: 105586/TN-3 Rev: Final B

1 Introduction .................................................................................................................................. A1
2 Lilleakers report ........................................................................................................................... A1
Report no: 105586/TN-3 Rev: Final B Page Ai

1 Introduction
In the project of building the leak frequency model, Lilleaker has built a databasis in excel format
with all HCR-data and developed additional data fields (based on the existing data fields), filters
and tools for data analysis. This appendix contains Lilleakers documentation of the HCR-data,
documentation of the developed databasis and also general considerations related to the data
fields in HCRD.
Lilleakerss report contains one main report and one Appendix. They are both given in the next
chapter.
2 Lilleakers report
Report no: 105586/TN-3 Rev: Final B Page A1

Report title: xxxxx Page: 1 of 26
Client: xxxxx Date: xx.xx.xx
Doc. no.: LA-xxxx-R-xxx Rev.: xx
HCR data for leak frequency model

Report title: HCR data for leak frequency model Page: 3 of 26
Client: LR Consulting Date: 06.11.2015
Doc. no.: LA-2010-R-064 Rev.: FINAL B
Table of contents:
1 Summary ....................................................................................................................................... 4
2 Introduction .................................................................................................................................. 7
2.1 Abbreviations ......................................................................................................................... 7
3 Process leak (leak scenarios) ........................................................................................................ 8
3.1 Process ................................................................................................................................... 8
3.2 Category ................................................................................................................................. 8
3.3 Severity .................................................................................................................................. 9
3.4 Hazardous area classification ................................................................................................. 9
3.5 System .................................................................................................................................... 9
3.6 Equipment .............................................................................................................................. 9
3.7 Major units ........................................................................................................................... 10
3.8 Blowout ................................................................................................................................ 10
3.9 Subsea leak........................................................................................................................... 11
4 Leak causes .................................................................................................................................. 13
4.1 Design cause ........................................................................................................................ 13
4.2 Procedural cause .................................................................................................................. 13
4.3 Equipment cause .................................................................................................................. 13
4.4 Operational cause ................................................................................................................. 13
5 Leak details ................................................................................................................................. 15
5.1 Hole size............................................................................................................................... 16
5.2 Initial leak rate ..................................................................................................................... 17
5.3 Duration ............................................................................................................................... 18
5.4 Actual pressure..................................................................................................................... 19
5.5 Quantity................................................................................................................................ 19
5.6 Inventory .............................................................................................................................. 20
5.7 Operational mode ................................................................................................................. 20
5.8 Gas detection ........................................................................................................................ 21
5.9 Other detection means.......................................................................................................... 21
6 Emergency reactions .................................................................................................................. 22
6.1 Shutdown ............................................................................................................................. 22
6.2 Blowdown ............................................................................................................................ 22
6.3 Deluge .................................................................................................................................. 22
6.4 Muster .................................................................................................................................. 22
6.5 Other .................................................................................................................................... 22
7 Population Data .......................................................................................................................... 24
8 Conclusion ................................................................................................................................... 25
9 References.................................................................................................................................... 26
APPENDIX
Appendix A HCR definitions
1 Summary
This report describes the contents of the HCR database [1] with the objective of using the
records of hydrocarbon leaks as a basis for making a process leak frequency model for use on
NCS.
It is important to have a common understanding of the definition of a process leak scenario.

Table 1-1 shows some categories in the HCR database that may be relevant for classifying a
process leak and evaluations of these.
Further classification of the process leaks based on their severity/potential/relevance for QRA.
Table 1-2 shows such fields and evaluation of these.
Each field in the database is described in Appendix A.
Table 1-1 Process leaks categories

HCR field # HCR data field HCR Description Comment
Installation type: FIXED, MOBILE,
Fixed installations have
SUBSEA. The installation may
2 CATEGORY equipment population
have a subsea satellite (recorded in
counts.
field 16 subsea)
This is the type of Hydrocarbon Equipment population
19 PROCESS released, i.e. NON-PROCESS OIL does not exist for non-
CONDENSATE GAS 2-PHASE process equipment
This field contains either a full
description of the system involved Some systems are not
28 SYSTEM or a Drilling or Well Operation relevant for process leak
activity description where scenarios.
appropriate.
This gives the full equipment item
Som equipment types
description. For Drilling/Well
32 EQUIPMENT are not relevant for
Operations activities (see item 28
process leak scenarios.
above) this will be left blank.
This field contains the Hazardous
Area Classification for the location Leaks in unclassified
of the incident, where 1 and 2 areas may not be
43 HAZ_CLASS
represent areas 1 and 2 relevant for process leak
respectively, and 3 represents scenarios.
unclassified.
HCR field # HCR data field HCR Description Comment

Module is not defined in
HCRD, however it is
stated:
3
3000m explosive
clouds [are] enough to
This contains the volume of the fill an entire module or
3
module involved, in m , and will deck area.
47 MOD_VOLUME Module volumes are
show NOT KNOWN where not
reported. sometimes reported to
be very small, maybe
inside confinements
such as separate rooms
(e.g. for pumps) or
under hood of turbines.
This is the isolatable hydrocarbon The inventory of a
inventory contained in the system, standard isolatable
53 INVENTORY in kgs. And will show NOT segment should be
KNOWN where not reported. significant. Many are
reported as very small.
May indicate that the
leak was not a process
leak. E.g. ROV detection
DETECTION_ or pressure drop may
58 Leak detected by other means
OTHER indicate subsea leak.
Subsea leaks are not
relevant for process leak
scenarios.
Table 1-2 Process leaks severity/potential/relevance for QRA leak frequency model
HCR field # HCR data field Description Comment
This shows the severity of the Leaks with minor
21 SEVERITY release as either MAJOR, severity may not be
SIGNIFICANT, or MINOR. relevant for QRAs.
Leaks with small
Amount of Hydrocarbon released in quantities released may
26 QUANTITY
kg not be relevant for
QRAs.
27 DURATION Duration of leak in minutes.
This is the hydraulic equivalent hole
size, deduced from d = 4A/p, in
mm. Where d is the diameter of the
hydraulic equivalent hole, A is the
This is in general an
cross-sectional area of the actual
2 unreliable data field. No
44 HOLE_DIAM hole in mm , and p is the wetted
hole sizes <1mm
perimeter of the actual hole in mm.
recorded before 2001.
It is important to note that N/A in
this field indicates that hole size is
not applicable to the mode of
release involved.
MAX_ This is the maximum allowable (Actual pressure > max
51
PRESSURE pressure of the system, in barg. pressure) may be a leak
52 ACT_ The actual (working) pressure at scenario of particular
HCR field # HCR data field Description Comment

PRESSURE time of incident, in barg. interest (rupture leaks).
This is the isolatable hydrocarbon The inventory of a
inventory contained in the system, standard isolatable
53 INVENTORY in kg/s. and will show NOT segment should be
KNOWN where not reported. significant. Many are
reported as very small.
60 EQUIP_CAUSE
Operation cause hard to
imagine for some
61 OP_CAUSE Leak causes.
equipment types, such
as piping.
62 PRO_CAUSE
The operational mode in the area at
62 OP_MODE
the time of release,
71 SHUTDOWN No action taken
72 BLOWDOWN indicates a less serious
73 DELUGE accident for loss of main
Emergency actions taken because safety function
74 CO2_HALON
of the leak
75 MUSTER
EMERACT_
76
OTHER
2 Introduction
The HCR database [1] includes 4561 leaks from the UK continental shelf from 3rd quarter
1992 to 1st quarter 2015. These data may act as a basis for building a process leak frequency
model. Since the model shall model process leaks, all leaks in the data basis may not be
relevant for this purpose and should be removed from the data basis.
QRAs usually models process leaks as leaks occurring spontaneously from a fully pressurized
process segment and is controlled by ESD and blowdown.
This document will discuss the entries in the HCR database and how they may be used as
basis for the leak frequency model.
The data in the HCR data base should be used with care. The sections below discuss some
findings in the data. Data found in this section is given in a separate excel worksheet [2]. This
report is structured to match the filters created in the worksheet.
Note that whenever leak counts are presented in this report, it is either based on the full set of
leak in the spreadsheet or an indicative subset called process leaks. For the final definition
of process leaks, see TN-4.
2.1 Abbreviations
Abbreviations used in this report are shown in Table 2-1. For abbreviations used in database
fieldnames, see appendix A.
Table 2-1 Abbreviations

Abbreviation. In full
HCR Hydrocarbon release
MISOF Modelling of ignition sources on offshore oil and gas facilities
ROV Remote Operated Vehicle
DNV Det Norske Veritas
N/A Not Applicable
3 Process leak (leak scenarios)

It is of critical importance that a user knows and understands what leaks are included in the
data set. NORSOK Z-013 section 7.4.4 describes process accidents as a specific category to
be analyzed in a QRA. Z-013 does however not define a process accident (it refers to the
HAZID), and therefore the QRA will define process accidents for each specific project or
client.
It is not within the scope of this report to establish a common or standard rule set for what to
include as a process leak in a QRA context. But since the project proposes leak frequencies
for use in QRA, it is important that a user of these frequencies understands what leaks
scenarios are included and what leak scenarios are not. This could be on a system level,
equipment level or even relate to causes or leak location. For example, are the following
process leaks that should be included in the recommended frequencies?
A leak that occurred outside a process area (non-hazardous area)

A leak from the flare system
A leak from the gas lift annulus through the wellhead
A leak during maintenance with a platform that is shut down
A leak that resulted from incompliance with procedures
These questions do not have correct yes/no answers, but for a user of generic leak frequencies
it is important that these battery limits are well defined and correctly understood.
From the description of incidents in the database, it is not always obvious whether a specific
incident should or should not be included in any given category of incidents. Rule sets will be
established, but the quality of the data and limitations to what is actually recorded means that
the number of incidents in any given leak category would be uncertain.
3.1 Process
This field refers to the fluid released, and non-process leaks should identify incidents that
are normally not considered process leaks in a QRA context.
3.2 Category
This field indidates installation type: FIXED, MOBILE, SUBSEA. The installation may have
a subsea satellite (recorded in field 16 subsea)
To what extent M and S type installations is part of scope and how these are reflected in
the population data is of interest.
3.3 Severity
The leak severity categories are defined in Appendix A. Note that severity is an automatic
evaluation based on other data fields. For the total data set considered, leaks are distributed on
the three severity categories as follows:
207 categorized as MAJOR (27 of these with hole size diameter D > 100)
2103 categorized as SIGNIFICANT (59 of these with hole size diameter D > 100)
2251 categorized as MINOR (27 of these with hole size diameter D > 100)
3.4 Hazardous area classification

The hazardous area classification for the location of the incident is included as field 43
HAZ_CLASS. Where leak is in unclassified area, the leak point is outside the process area.
The relevance of such incidents to process area risk analysis can be discussed. The data set
contains 147 process fluid leaks where the area is categorized as unclassified, so this is not a
large fraction of the incidents.
The information in this data field may not always reliable. For example, some of the subsea
leaks are recorded in zone 2.
3.5 System
All leaks are assigned to a system. This field contains either a full description of the system
involved or a Drilling or Well Operation activity description where appropriate.
Some systems are obviously relevant when it comes to defining a process leak, such as
separation or compression. Others are less obvious, such as releases from the drain or
drilling systems. Which systems are relevant for the process leak frequencies to be
established?
Leaks from the open and closed drain system could be hard to interpret. The hydrocarbons
have come from process equipment via the drain system. There are 198 leaks from drain or
open drain systems of which 112 are minor. For example, there are three leaks from pressure
vessels (equipment type) in the open drain system. It is believed that the pressure vessel is
part of another system, while the released fluid is from the drain system. (Drain tank should
normally not be defined as a pressure vessel).
3.6 Equipment
This field gives the leaking equipment description. Most leaks are assigned to an equipment
type (some are N/A). Note that sometimes equipment type and system type appear to be in
conflict.
Equipment that is generally not considered process leaks includes categories such as riser
and BOP.
For Drilling/Well Operations activities this will be left blank.

Piping
There are 1144 leaks from piping. Of these, 188 incidents have equipment cause NONE.
Operational cause is LEFTOPEN, OPENED or IMPROPOP for 93 of these. Of these
93, 12 have hole diameter N/A, 17 have hole diameter > 100mm and 17 have diameter sizes
in the range 1 to 3. See chapter 4.4 for further discussions.
An important point is that the fraction of large hole diameters is quite different:
For those 93 leaks with no equipment cause and operational cause as above, 13% have
hole diameter > 100mm.
For the remaining 956 leaks (with equipment cause NONE), 0.9% (9 incidents) have
hole diameter > 100 mm. For these 9 incidents, duration is anything from very short (5
seconds) to very long (8 days).
It may well be that the operational piping leaks with D > 100 and D = N/A are similar
incidents. For the SHLFM [3], N/A are discarded (D< 1 mm) while D > 100 certainly
contributes to the large leak category. Further, there are likely to be many similar incidents in
the 1 4 range as well (see chapter 4.4).
Discussion: The N/A incidents are likely to be less severe than the > 100 incidents. This
should be further addressed in order to justify omission of incidents with hole size N/A.
3.7 Major units

The definition of major equipment (which includes e.g. Pressure vessels as separators) in
HCRD is as follows (see appendix A):
Each item comprises the item of equipment itself, but excluding all valves, piping, flanges,
instruments and fittings beyond the first flange and excluding the first flange itself.
The definition of Instruments in HCRD:

One Instrument could comprise the instrument itself, plus up to 2 valves, up to 4 flanges, 1 fitting, and
associated small bore piping (1"or less).
It is Lilleakers understanding that leaks from instrument connections on major equipment are
recorded as leaks from the major equipment:
The first flange does not exist for an instrument connection because this is included
in the definition of the instrument it self
The leak data seem to suggest that this is the case: several recorded hole sizes of 0.5,
1 and 2 may correspond to rupture of instrument connections.
3.8 Blowout
Blowout is not a category in the database. 11 leaks with system containing well or
drilling have duration of 24 hours or more. One incident seems to be a blowout (Year 2012,
number 125), the remaining are different well leaks scenarios that were not detected by gas
detectors (one exception). One is detected as a fire Flame.
24 UK blowouts and well releases are included in the Sintef offshore blowout database for
this period 1992-2015. Of these, 4 are releases are from X-mas tree or wellheads.
Table 2: HCRD incidents x-mas tree or wellhead that are found in the blowout database
Sintef Offshore Category (Sintef Offshore blowout database)
HCR ID
blowout database ID
Limited surface flow before the secondary barrier
1994-1995-25 490
was activated
Limited surface flow before the secondary barrier
1995-1996-146 497
was activated
1996-1997-99 492 Totally uncontrolled flow, from a deep zone
2011-2012-125 626 Totally uncontrolled flow, from a deep zone
One incident from UKCS for the period 1992-2015 and none from the NCS are included in
the estimate for Blowout and Well release frequencies for producing wells for use on NCS, as
reported in the annual LR consulting report (Two incidents from the UKCS in 1988 and 1989,
respectively, are included.)
Reference is made to the latest annual report: Blowout and well release frequencies based on
SINTEF offshore blowout database 2014 Report no: 19101001-8/2015/R3 Rev: Final, March
17th 2015 [4], tables 4.1 to 4.4.
The one incident is a well release from 2007 and has ID 596 in the Sintef Offshore blowout
database. This is a subsea release and not relevant for the Leak frequency model.
Description of ID596 from the Sintef Offshore blowout database:
Wells Incident -<...>Incident reported by field standby vessel "Putford Artemis". Vessel
reportes bubbles coming to surface with a 10m dispersion radius at location of <...>subsea
wellhead structure.<...> responded as contractedd operator through a sequence of shut
downs to determine the hydrocarbon gas release was from the B1 (B9) well. The well was
shut in and the gas release stopped. The well remains shut in and will require inspection of
the structure to ascertain the causef cause of the gas release.
In Lilleakers opinion, no adjustment has to be made for the Process leak frequency model
based on events included in the blowout and well release frequencies for producing wells.
3.9 Subsea leak

Subsea leaks should be not included in the data set. It is not straight-forward to identify these
leaks from the HCR database. (There are some examples of subsea leaks that have been
included in the MISOF data set.)
If one or more of the following is true, the leak should be considered a subsea leak:
Category = S1
System contains SUBSEA
Equipment contains SUBSEA
Detection other = ROV
53 leaks are identified this way as subsea leaks. Most likely, there are more subsea leaks in
the dataset after this exercise. Note that the fields ventilation, no of sides and mod
volume are typically set to NOT KNOWN for these leaks, while air changes seems to be
not known in every case. So these fields may also be an indicator for a subsea leak.
Another indicator for a subsea leak may be a leak with long duration. 56 different leaks not
detected by filters above with non process = (empty) have a duration of 24 hours or more.
Of these leaks were 36 leaks from systems that may be subsea systems.
For subsea leaks, the field HAZ_CLASS should be unclassified, but this is not the current
practice in the database. It seems like some subsea leaks may have HAZ_CLASS=2, which
is the case for subsea wells.
1
See Appendix A for description of the different categories in the HCR database
4 Leak causes
It may be of interest to look into what caused a leak. As we understand, industry practice for
process leak analyses has been to consider all causes as relevant. This may not be the case for
other parts of the QRA such as collision (way-point at installation) riser leak and blowouts
(external causes), and dropped objects (lifting restrictions).
Anyway, it is of interest to look into what caused the incidents that pass a set of other criteria.
When a particular type of equipment is analyzed, it is important to know whether the fault is
an equipment fault or not. An example here that is further discussed is piping leaks that have
no equipment failure. These have mostly operational causes. It may not be a productive to mix
these incidents with piping leaks caused by corrosion or mechanical failure.
4.1 Design cause

This field in the database indicate that the failure was related to design.
Of the total of 4561 leaks, 629 are recorded to have a design cause.
Of 2758 process leaks, 373 are recorded to have a design cause.
4.2 Procedural cause

This field in the database indicated that the failure was related to procedures (both non-
compliance and deficient procedures).
Of the total of 4561 leaks, 1070 are recorded to have a procedural cause.
Of 2758 process leaks, 545 are recorded to have a procedural cause.
4.3 Equipment cause

This field in the database indicated that the failure was related to the equipment itself such as
corrosion, erosion mechanical fatigue.
Of the total of 4561 leaks, 2895 are recorded to have an equipment cause.
Of 2758 process leaks, 1881 are recorded to have an equipment cause.
4.4 Operational cause

704 leaks have operational cause OPENED LEFT OPEN or IMPROPOP while the
equipment cause is NONE. 58 of these have hole size > 100. This is 50% of the leaks with
D > 100. Equipment type varies, but many are piping, flange or valve. Is there any good
reason to scale these events with the number of flanges, valves or piping length except that all
of these could be good indirect measures of activities that could involve all types of mistakes?
For the leaks with operational causes as listed above, equipment type for most of the incidents
is listed as piping, flange or valve. The question to ask is whether this categorization to some
extent is arbitrary. Say a valve is opened and gas is released as a consequence. Could it be that
in this case the operator has a difficult task to decide if the equipment type is the valve that
was opened, the piping the gas was released through, or the flange at the end of the piping?
And the hole size, would that be the diameter of the piping (even if other restrictions might
exist)? Or would some operators perhaps record N/A for the hole size for the very same event.
Physically, piping cannot be opened to cause a leak, since piping is a simply a physical
barrier. A valve may be opened, and a flange could be opened as well. This could be
important for several reasons. If a pipe is routed through an area and there are no flanges or
connections of any kind, what is the leak frequency? Opened is not really an option. The
relatively large number of leaks (with large hole diameters) due to operational causes would
not be applicable in this case.
Assigning the leaks caused by operational mistakes to equipment type (such as piping) could
potentially be misleading and lead to incorrect focus and decisions when it comes to risk
assessments or mitigation means. This does not mean that the population of valves and
flanges cannot be a reasonably good indicator for the leak frequency also for operational
leaks.
5 Leak details
Leak details include the quantity and duration and inventory of the leaks. Actual pressure and
maximum pressure are included here as well, in addition to the recorded hole size. Finally,
operational mode is included. This is relevant information for describing the consequence of a
leak. The following data fields are relevant in this context.
Hole size
Actual pressure
Max pressure
Quantity
Duration
Inventory
Operational mode
Hazardous class
Severity
The rules for which leaks are reportable are very strict: Leaks with rate > 1 kg per hour (gas)
or 5 kg per day (liquid) are reportable. Many small leaks may not be of interest for QRAs.
The flowchart for deciding whether a leak is reportable or not is shown in Figure 5-1.
Figure 5-1 Ratio between Flowchart on Reportability of Hydrocarbon Releases
5.1 Hole size

In HCR, Hole sizes are the hydraulic equivalent hole size, deduced from d = 4A/p, in mm.
Where d is the diameter of the hydraulic equivalent hole, A is the cross-sectional area of the
actual hole in mm2, and p is the wetted perimeter of the actual hole in mm.
It is important to note, that those releases with a hole size labelled N/A are special cases
where the release rate is not applicable to the mode of release (e.g. open topped vessels such
as shale shakers, or where carry-over of hydrocarbons from one system to another was
involved). All such releases were classified by inspection of the amount released only. Hole
sizes less than 1 mm are set to 1 before 2001. It is also debatable how easy it is to be
consistent when measuring the hole diameter. The hole may be everything from a full rupture,
to a small fracture or a poor fitted flange coupling.
In all, there are 160 leaks with hole size N/A. 111 of these have equipment cause NONE. 4
of these have severity MAJOR.
There are 113 leaks with hole size > 100. 83 of the latter have operational cause NONE. 17
of these have severity MAJOR.
5.2 Initial leak rate

Leak rate is not reported in the HCRD. The graph in Figure 5-2 shows the ratio between
calculated initial rate and average rate. For almost 3000 leaks, the initial rate is between 75%
and 200% of the average leak rate. For about 500 incidents, the initial rate is between 2 times
and 10 times the average rate. For the remaining 500 incidents, the initial leak rate is more
than a factor 10 higher than the average rate.
For about 500 leaks, the calculated initial rate is significantly less than the average rate.
Except if the leak rate was increasing over time, the calculated initial rate is too low for these
leaks. Most of these incidents are categorized as Zero pressure leaks in [3]. For 287 leaks
with average rate ten times or more higher than the calculated initial rate, 38 have initial rate
exceeding 1 kg/s.
For a few leaks, the two values are very different, indicating that something is incorrect. The
initial leak rate is calculated with the method used in Standardised Hydrocarbon Leak
Frequencies [3].
100000
10000
1000
Initial leak rate / average leak rate
100
10
0.1
0.01
0 500 1000 1500 2000 2500 3000 3500 4000 4500 5000
Number of leaks
Figure 5-2 Ratio between calculated initial rate and average rate
In Figure 5-2, leak rates for hole sizes > 100 mm are calculated based on a hole diameter of
110mm. An alternative calculation with 220 mm hole size was performed. The resulting graph
is virtually identical with the one shown.
5.3 Duration
Normally, process leaks will have durations of more than 1 minute and less than one hour due
to the size of isolatable segments of the process plant and safety systems such as blowdown.
Most leaks in the HCR database are within this category.
Leaks with very short duration would normally be leaks from a very limited inventory. It
seems that leaks with very long duration are in many cases not really process leaks but may
for instance be subsea leaks.
55 leaks have duration less than 5 seconds

151 leaks have duration 15 seconds or less
714 leaks have duration 1 minute or less
3147 leaks have duration between 1 minute and 60 minutes
700 leaks have duration 1 hour or more
81 leaks have duration 24 hours or more
16 leaks have duration 1 week or more
5.4 Actual pressure

Actual pressure is a very central data field for the current leak frequency model. The pressure
is used for calculating initial leak rates and for classification of the leaks.
For all but one leak, an actual pressure is recorded. For two leaks, the actual pressure is
slightly less than zero. For 334 leaks, the actual pressure is less than or equal to 0.01 barg.
1000
100
10
Actual pressure
0.1
0.01
0 500 1000 1500 2000 2500 3000 3500 4000 4500 5000
Number of leaks
Figure 5-3: Distribution for actual pressure
5.5 Quantity
Leaks are registered with the amount of hydrocarbon released; this field is called Quantity in
the database.
For about 50% of the leaks in the HCR database, the released quantity is less than 10 kg. The
relevance of these leaks should be debated. Below the number of leaks is shown for different
quantity categories. The total number of leaks is 4561.
Quantity < 1 kg: 1095 leak

Quantity < 10 kg: 2358 leaks
Quantity < 100 kg: 3628 leaks
Quantity > 10 000 kg: 42 leaks
Quantity> 50 000 kg: 13 leaks
Quantity > 100 000 kg: 7 leaks
(Of these 7 leaks, 1 is apparently a blowout, 1 flaring, 1 storage tank, 1 pipeline, 1 subsea, 1
manifold -with duration 6 days , 1 with duration 73 days-export oil, piping, mech. ventilated
area of unknown volume.)
5.6 Inventory
Of the total 4561 leaks
1092 are reported with inventory < 100 kg
808 with inventory 100-1000 kg
426 with inventory > 10000 kg
1417 with inventory NOT KNOWN
Of 2758 process leaks

636 are reported with inventory < 100 kg
287 with inventory > 10000 kg
682 with inventory NOT KNOWN
Incidents with inventory not known seem to include all types of systems, and not restricted to
systems with inventory that is hard to define such as wells.
For 69 leaks, inventory is reported to zero and in 233 cases less than 1 kg. Again, these leaks
are from all kinds of systems. In some cases, inventory might have been set to zero rather than
not known. For some leaks, the system might have been empty when intrusive maintenance
is initiated. The gas or oil might then come from faulty isolation from a neighboring segment.
5.7 Operational mode

11 different operational modes are recorded.
Of 4561 leaks, 2495 are recorded during normal operation.

Of 2758 process leaks, 1692 are during normal operation.
5.8 Gas detection

This shows whether a GAS detector was activated.
Of the total of 4561 leaks, 1712 are recorded with gas detection
Of 2758 process leaks, 1111 are recorded with gas detection.
5.9 Other detection means

20 leaks are detected by use of ROV. Not all of these are easily identified as subsea leaks.
There are also 32 leaks detected by pressure change. Some of these appear to be subsea leaks
as well.
6 Emergency reactions
Emergency reactions include actions such as shutdown and blowdown, but also deluge, and
muster. These may give useful additional information on the incident. For example if no
shutdown or blowdown was initiated this is an incident that has a development deviating from
what is commonly modelled in a QRA.
6.1 Shutdown
This field signifies that shutdown took place, either automatically or manually initiated.
Of the total of 4561 leaks, 3020 are recorded to have been shut down (manual or
automatic)
Of 2758 process leaks, 1938 are recorded to have been shut down (manual or
automatic)
6.2 Blowdown
This field signifies that blowdown took place, either automatically or manually initiated
Of the total of 4561 leaks, 1563 are recorded with blowdown initiated (manual or
automatic).
Of 2758 process leaks, 1182 are recorded with blowdown initiated (manual or
automatic).
6.3 Deluge
This field signifies that deluge took place, either automatically or manually initiated
Of the total of 4561 leaks, 122 are recorded with deluge initiated (manual or
automatic).
Of 2758 process leaks, 72 are recorded with deluge initiated
6.4 Muster
This field signifies that a muster took place at stations or at the lifeboats.
Of the total of 4561 leaks, 1225 are recorded with mustering initiated (at life boats or
at stations).
Of 2758 process leaks, 713 are recorded with mustering initiated (at life boats or at
stations).
6.5 Other
If any other emergency action was taken during the incident, but was not adequately covered
by any of the previous fields, it is recorded in this field.
Of the total of 4561 leaks, 1225 are recorded with other emergency reaction initiated.
Of 2758 process leaks, 713 are recorded with mustering initiated
7 Population Data
The population data should be used with care in this study. Not all systems with recorded
leaks have population data. It is for instance recorded leaks in the flare systems, but the
population data (on equipment) does not contain any data for this system. The population has
also been more or less constant since 2006, indicating update problems. The population data
does not contain the same amount of information as the leak data. Therefore, it is difficult to
use the same filters for the population data as for the leak data.
3164 leaks are registered with population data (equipment type). This means that 31 % of the
leaks are in systems that does not contain population data. The table below shows the
percentage of leaks registered in each severity category. The leaks with population data seem
to have similar distribution among the severity categories.
Table 7-1 leaks in different severity categories, all leaks and leaks with population data
Percentage of leaks registered with
Severity Percentage of all leaks
population data
Major 5% 5%
Significant 46 % 48 %
Minor 49 % 47 %
8 Conclusion
This projects intention is to use the HCR database for establishing generic frequencies for
process leaks. To do this, process leaks that match the purpose for these generic frequencies
must be identified. Many data fields in the HCR database [1] can be used for categorization
of incidents as a process leak scenario or not.
Leak scenarios recorded in HCR may, however, differ from what is usually modelled in
QRAs. The frequency assigned to the scenarios usually modelled in QRAs must be based on
carefully selected subset of the database.
9 References
[1] Hydrocarbon Releases System, https://www.hse.gov.uk/hcr3/index.asp.
[2] HCR, data, Excel Workbook.
[3] Offshore QRA - Standardised Hydrocarbon Leak Frequencies, Report No 2008-
1768/1241Y35-14, Rev. 1 2009.
[4] LR Consulting, Blowout and well release frequencies based on SINTEF offshore blowout
database 2014 Report no: 19101001-8/2015/R3 Rev: Final, March 17th 2015,, 2015.
Appendix B
Extracted data from HCRD and

estimated leak frequencies

1 Introduction ...................................................................................................................................B1
2 Relevant process leaks fed through process system ........................................................................B2
2.1 Q3 1992 Q1 2015 ............................................................................................................B2
2.2 Q1 2001 Q1 2015 ............................................................................................................B5
3 Relevant process leaks fed through utility system ...........................................................................B8
3.1 Q3 1992 Q1 2015 ............................................................................................................B8
3.2 Q1 2001 Q1 2015 ..........................................................................................................B11
4 Relevant process leaks fed through process system or utility system .............................................B14
4.1 Q3 1992 Q1 2015 ..........................................................................................................B14
4.2 Q1 2001 Q1 2015 ..........................................................................................................B17
5 Relevant leaks from well system ...................................................................................................B20
5.1 Q3 1992 Q1 2015 ..........................................................................................................B20
5.2 Q1 2001 Q1 2015 ..........................................................................................................B21
6 Exposure data ..............................................................................................................................B22
6.1 Process equipment ............................................................................................................B22
6.2 Well head ..........................................................................................................................B23
7 Estimated leak frequencies based on HCRD .................................................................................B24
7.1 Q3 1992 Q1 2015 ..........................................................................................................B24
7.2 Q1 2001 Q1 2015 ..........................................................................................................B27
Report no: 105586/TN-3 Rev: Final B Page Bi

1 Introduction
This Appendix contains detailed data extracted from HCRD. The data is used as basis for
estimating leak frequencies (per equipment year) for the defined leak types covered by the model.
Calculated leak frequencies based on HCR-data are given in detail in this appendix.
Report no: 105586/TN-3 Rev: Final B Page B1

2 Relevant process leaks fed through process system
2.1 Q3 1992 Q1 2015
Table 2.1 - Relevant process incidents fed through process systems for the period Q3 1992 - Q1 2015. It is distinguished on leaked quantity, initial
pressure, and on hole sizes 1 mm, >1 mm and hole sizes recorded as N/A

Table 2.2 - Relevant process incidents fed through process systems for the period Q3 1992 - Q1 2015. The leaks are categorized into the defined leak
scenarios for the model (see TN-4). It is also distinguished on hole sizes 1 mm, >1 mm or N/A

scenarios for the model (see TN-4). It is also distinguished on hole sizes 1 mm, >1 mm or N/A. All equipment size categories of actuated and manual
valves, standard flange, steel pipe and shell and tube side heat exchangers are grouped together

2.2 Q1 2001 Q1 2015
Table 2.4 - Relevant process incidents fed through process systems for the period Q1 2001 - Q1 2015. It is distinguished on leaked quantity, initial



3 Relevant process leaks fed through utility system
3.1 Q3 1992 Q1 2015
Table 3.1 - Relevant process incidents fed through utility systems for the period Q3 1992 - Q1 2015. It is distinguished on leaked quantity, initial

Table 3.2 - Relevant process incidents fed through utility systems for the period Q3 1992 - Q1 2015. The leaks are categorized into the defined leak


3.2 Q1 2001 Q1 2015
Table 3.4 - Relevant process incidents fed through utility systems for the period Q1 2001 - Q1 2015. It is distinguished on leaked quantity, initial



4 Relevant process leaks fed through process system or utility system
4.1 Q3 1992 Q1 2015
Table 4.1 - Relevant process incidents fed through process or utility systems for the period Q3 1992 - Q1 2015. It is distinguished on leaked quantity,
initial pressure, and on hole sizes 1 mm, >1 mm and hole sizes recorded as N/A

Table 4.2 - Relevant process incidents fed through process or utility systems for the period Q3 1992 - Q1 2015. The frequencies are categorized into the
defined leak scenarios for the model (see TN-4). It is also distinguished on hole sizes 1 mm, >1 mm or N/A

Table 4.3 - Relevant process incidents fed through process or utility systems for the period Q3 1992 - Q1 2015. The frequencies are categorized into the
defined leak scenarios for the model (see TN-4). It is also distinguished on hole sizes 1 mm, >1 mm or N/A. All equipment size categories of valves,
standard flange, steel pipe and shell and tube side heat exchangers are grouped together

4.2 Q1 2001 Q1 2015
Table 4.4 - Relevant process incidents fed through process or utility systems for the period Q1 2001 - Q1 2015. It is distinguished on leaked quantity,
initial pressure, and on hole sizes 1 mm, >1 mm and hole sizes recorded as N/A

Table 4.5 - Relevant process incidents fed through process or utility systems for the period Q1 2001 - Q1 2015. The leaks are categorized into the
defined leak scenarios for the model (see TN-4). It is also distinguished on hole sizes 1 mm, >1 mm or N/A

Table 4.6 - Relevant process incidents fed through process or utility systems for the period Q1 2001 - Q1 2015. The leaks are categorized into the
defined leak scenarios for the model (see TN-4). It is also distinguished on hole sizes 1 mm, >1 mm or N/A. All equipment size categories of actuated

5 Relevant leaks from well system
5.1 Q3 1992 Q1 2015
Table 5.1 - Relevant incidents from well systems for the period Q3 1992 - Q1 2015. It is distinguished on leaked quantity, initial pressure, and on hole
sizes 1 mm, >1 mm and hole sizes recorded as N/A
Table 5.2 - Relevant process from well systems for the period Q3 1992 - Q1 2015. The leaks are categorized into the defined leak scenarios for the
model (see TN-4). It is also distinguished on hole sizes 1 mm, >1 mm or N/A

5.2 Q1 2001 Q1 2015
Table 5.3 - Relevant incidents from well systems for the period Q1 2001 - Q1 2015. It is distinguished on leaked quantity, initial pressure, and on hole
sizes 1 mm, >1 mm and hole sizes recorded as N/A
Table 5.4 - Relevant process from well systems for the period Q1 2001 - Q1 2015. The leaks are categorized into the defined leak scenarios for the
model (see TN-4). It is also distinguished on hole sizes 1 mm, >1 mm or N/A

6 Exposure data
6.1 Process equipment
Table 6.1 - Exposure data for process equipment as part of process systems

Table 6.2 - Exposure data for process equipment as part of process systems. All equipment size
categories of actuated and manual valves, standard flange, steel pipe and shell and tube side heat
exchangers are grouped together
6.2 Well head

Table 6.3 - Exposure data for well heads

7 Estimated leak frequencies based on HCRD
7.1 Q3 1992 Q1 2015
Table 7.1 - Estimated process leak frequency based on HCRD for the period Q3 1992 - Q1 2015. It is distinguished on leaked quantity, initial pressure,
and on hole sizes 1 mm, >1 mm and hole sizes recorded as N/A

Table 7.2 - Estimated process leak frequency based on HCRD for the period Q3 1992 - Q1 2015. The leaks are categorized into the defined leak

Table 7.3 - Estimated process leak frequency based on HCRD for the period Q3 1992 - Q1 2015. The frequencies are categorized into the defined leak

7.2 Q1 2001 Q1 2015
Table 7.4 - Estimated process leak frequency based on HCRD for the period Q1 2001 - Q1 2015. It is distinguished on leaked quantity, initial pressure,
and on hole sizes 1 mm, >1 mm and hole sizes recorded as N/A

Table 7.5 - Estimated process leak frequency based on HCRD for the period Q1 2001 - Q1 2015. The leaks are categorized into the defined leak

Table 7.6 - Estimated process leak frequency based on HCRD for the period Q1 2001 - Q1 2015. The frequencies are categorized into the defined leak

Appendix C
Complementary cumulative
hole size distributions and leak
rate distributions based on
HCRD

1 Introduction .................................................................................................................................. C1
2 Complementary cumulative hole size distributions based on HCRD .............................................. C2
2.1 Log-log plots ...................................................................................................................... C2
2.2 Linear plots ....................................................................................................................... C12
3 Complementary cumulative leak rate distributions based on HCRD............................................. C22
3.1 Log-log plots .................................................................................................................... C22
3.2 Linear plots ....................................................................................................................... C32
Report no: 105586/TN-3 Rev: Final B Page Ci

1 Introduction
This appendix gives complementary cumulative hole size distributions for all equipment types in
HCRD. The hole size distributions are plotted in log-log in Chapter 2.1 and linear plots in Chapter
2.2. Correspondingly complementary cumulative leak rate distributions are given in log-log in
Chapter 3.1 and linear scale in Chapter 3.2. Filter 1 is used to extract incidents as basis for hole
size distributions. Filter 2 and 3 are used to analyse the effect of including process leaks fed
through utility systems, and including incidents recorded with total released quantity <10 (Margi-
nal leaks) and leaks with initial pressure <0.01 barg. Filter 4 is defined to produce estimated leak
rate distributions based on the same type of incidents as the leak rate distributions based on NCS
data are based on (see TN-2). Note that there are uncertainties related to both the hole size distri-
butions and the leak rate distributions (see TN-3). Filters extracting incidents from the period Q3
1992- Q1 2015 are denoted a, while filters from the period Q1 2001- Q1 2015, are denoted
b. The filters are defined in the table below. See also TN-3.
Table 1.1 - Filters used to extract incidents (hole sizes) as basis for recorded hole size distributions
and estimated leak rate distributions based on HCRD
Filter Description
Filter 1 All relevant process leak and well system leak incidents in the HCR-data as
defined in Chapter 2.1 and 2.3 (in TN-3) are included, except:
This filter is put as basis for hole size distributions in the model development
Filter 2 All relevant process leak incidents in the HCR-data as defined in in Chapter 2.1 (in
TN-3), relevant utility leaks as defined in Chapter 2.2 (in TN-3) and relevant well
releases as defined in 2.3 (in TN-3) are included, except:
This filter is defined to analyse the effect of including process leaks fed through
utility systems and well systems as basis for hole size distributions.
Filter 3 All relevant process leak incidents in the HCR-data as defined in in Chapter 2.1 (in
This filter is defined to also analyse the effect of including incidents recorded with
pressure <0.01 barg, and incidents recorded with total released quantity <10 kg.
Filter 4 All relevant process leak incidents in the HCR-data as defined in Chapter 2.1 (in
Incidents recorded with initial leak rate <0.1 kg/s
This filter is defined to establish leak rate distributions based on the same type of
incidents as the leak rate distributions based on NCS data are based on
Report no: 105586/TN-3 Rev: Final B Page C1

2 Complementary cumulative hole size distributions
based on HCRD
The complementary cumulative hole size distributions are based on recorded hole sizes in HCRD.
2.1 Log-log plots

2.2 Linear plots

3 Complementary cumulative leak rate distributions
based on HCRD
The complementary cumulative leak rate distributions are calculated based on hole size and
available process conditions in HCRD.
3.1 Log-log plots

3.2 Linear plots

Working together
for a safer world
Leak scenarios
TN-4
Technical note for:


Date: 18 March 2016
1 Introduction .....................................................................................................................................1
2 Definition of leak scenarios considered in a QRA .............................................................................1
2.1.1 Blowout ............................................................................................................................. 3
2.1.2 Well release ........................................................................................................................ 3
2.2 Process leak ...........................................................................................................................3
2.3 Utility leak .............................................................................................................................4
2.4 Riser leak ...............................................................................................................................4
2.5 Pipeline leak ..........................................................................................................................4
2.6 Storage tank leak ..................................................................................................................4
3 Leak scenarios covered by the model ...............................................................................................4
4 Leak scenarios modelled in QRAs .....................................................................................................7
4.1 Process leaks .........................................................................................................................8
4.2 Producing well leaks ..............................................................................................................9
4.3 Gas lift well leaks...................................................................................................................9
4.4 Initial leak rate boundary for leaks considered in a QRA.........................................................9
4.5 Significant leaks vs. Marginal leaks ......................................................................................12
5 Leak scenarios not covered by the model .......................................................................................13
5.1 Zero pressure leaks ..............................................................................................................14
5.2 Limited leaks .......................................................................................................................15
5.3 Vent leaks within design specification..................................................................................17
6 References .....................................................................................................................................18

1 Introduction
NORSOK Z-013 Chapter 7.4.4, Ref. /1/, describes process accidents as a specific category to be
analysed in a QRA, but Z-013 does not define a process accident (it refers to the HAZID).
Although industry practice is quite consistent, it is observed that there may exist differences in
the industry with regard to how process accidents are defined by various stakeholders.
To improve consistent application throughout industry, the leak scenarios that are normally
considered in Quantitative Risk Analysis (QRA) for offshore oil and gas facilities are defined in this
technical note (TN). A leak should fall into only one of the defined leak categories.
These definitions form the basis for leak scenarios included in the leak frequency model.
expressions.
2 Definition of leak scenarios considered in a QRA

This chapter presents the definition of leak scenarios normally included in a QRA.
A leak from the well system is defined as a leak where a hydrocarbon fluid is released through a
hole located in the well system (classified as either producing well, gas lift well or injection well in
PLOFAM). Leaks from the well system are divided into blowouts and well releases. These releases
are further divided into releases during drilling, well operations and releases during normal
operation. The following barriers define the process system and the well system in this context:
Process system and producing well or Injection well:
o Production Wing Valve (PWV, see Figure 3.1)
Process system and gas lift annulus:
o Different configurations are applied in industry (see for instance Ref. /2/), such as
annulus wing valve(AWV)/ESV towards topside and Annulus safety valve (ASV) /Annulus
safety check valve (ASCV) towards well.

AWV
Figure 2.1 - Well barriers schematic for a standard producing well (left) (taken from Ref. /3/
Figure 24), and a gas lift well (right) (taken from Ref. /2/ Figure 1). An item that is coloured red (e.g.
Hydraulic Master Valve (HMV)) indicate that the item is a secondary barrier element towards the
formation fluid, whilst a blue coloured item (e.g. Downhole Safety Valve (DHSV)) indicate that the
piece of equipment is a primary barrier element towards the formation fluid

2.1.1 Blowout
The definition of a blowout, which is in accordance with Ref. /4/, is as follows:
An incident where formations fluid flows out of the well or between formations layers after all
the predefined technical well barriers, or the activation of them, have failed.
Leaks that occur in relation to well intervention operations (e.g. wireline or coiled tubing) are
defined to be a blowout if the leak occurs at the reservoir side of the BOP, or if the BOP does not
close. Furthermore, leaks that occur during normal operation are defined to be a blowout if the
DHSV fails.
2.1.2 Well release
The definition of a well release, which is in accordance with Ref. /4/, is as follows:
An incident where hydrocarbons flow from the well, at some point where flow was not intended
and the flow was stopped by use of the barrier system that was available for the well at the time
of the incident.
Leaks that occur in relation to well intervention operations (e.g. wireline or coiled tubing) are
defined to be a well release if the leak occurs on the platform side of the wireline or coiled tubing
BOP, and the BOP is functioning as intended.
During normal production the following (topside) well releases can occur from the wellhead or
x-mas tree:
Producing well/Injection well: Release of fluid from the inventory in the production/injection
line between the Downhole safety valve (DHSV, see Figure 1) and the PWV
Gas lift well: Release of fluid from the annulus inventory between the Annulus Safety Valve
(ASV) and the barrier towards annulus topside (e.g. EV, AWV) (see Figure 2.1). In cases
where no ASV is present, the entire inventory in the gas lift annulus to the GLV may be
released (see Figure 2.1). Incidents causing a topside leak from the gas lift annulus where the
ASV and the ASCV between gas lift annulus and production string fail, are considered to be
a blowout, and therefore not considered as a gas lift well leak. Producing well/injection
well/gas lifted well: Release of hydrocarbon fluid from annuli that are not used for gas lift.
2.2 Process leak

A process leak is defined as a scenario where fluid from a process system is unintentionally
released to the surroundings. The fuel gas system is regarded as part of the process system.
According to this definition, a leak point in a different system than the process system that is
being fed by fluid from the process system is a process leak. Examples of such incidents are leaks
in utility systems (drain systems, vent system and flare systems). However, if process fluid is
released through a hole located in the well system during normal production, the leak is defined
to be a well leak (producing well leak or gas lift well leak).
An unintentional release of fluid from a process system to flare tips and atmospheric vents is in
general not defined as a process leak except in cases where the release rate exceeds the design
specification and poses a fire and explosion hazard to equipment, structures or personnel.
The interface towards the well system is defined in Chapter 2.

2.3 Utility leak
A utility leak is defined as a scenario where fluid from a utility system is unintentionally released
to the surroundings. There is no communication with the main process system at any point in
time during the time span of the leak. Examples of utility system leaks are:
Leak of hydraulic oil from a hydraulic system
Leak of TEG from a TEG system
Leak of diesel from fuel supply system to a power generation system
Leak for produced water from a produced water system.
Note that the list is not exhaustive.
2.4 Riser leak

A riser leak is a leak that is fed from the riser and/or pipeline system. The leak point is in the riser
itself or equipment (e.g. flange, valve) associated with the riser. The boundary between the riser
system and the process system is the topside riser ESD valve.
The boundary between the riser system and pipeline system is a subsea valve, the flange conne-
ction with the pipeline or a spool piece at the sea bed.
2.5 Pipeline leak

A pipeline leak is a leak that is fed from the pipeline, with or without contribution from riser
depending on subsea isolation valve or not is present and functioning. The leak point is in the
pipeline itself or equipment (e.g. flange, valve) associated with the pipeline.
The boundary between the riser system and pipeline system is a subsea valve, the flange
connection with the pipeline or a spool piece at the sea bed.
If there is no subsea isolation valve segregating the riser and the pipeline, the boundary between
the pipeline/riser system and the process system is the topside riser ESD valve.
2.6 Storage tank leak

A storage leak is defined as a scenario where fluid from the storage tank system is unintentionally
released to the surroundings.
3 Leak scenarios covered by the model

Chapter 2 defined leak scenarios normally considered in a QRA. Not all of these scenarios are
covered by the leak frequency model (PLOFAM). This chapter specifies the leak scenarios covered
by the model, which is used to select relevant incidents as basis for frequency estimation and
estimation of hole size distributions for the leak scenarios suggested used for modelling in QRAs,
given in Chapter 4.
The leak frequency model covers topside hydrocarbon process leaks at NCS. Hence, the leak
scenarios covered by the leak frequency model are in accordance with the definition of a process
leak as defined in Chapter 2.2. In addition topside leaks from the well system occurring during
normal production, as defined in Chapter 2.1.2, are covered by the model. Other leak scenarios,
such as leaks from utility systems fed from utility systems (for example diesel from diesel tanks
and MEG from MEG-system) are not included.
The leak scenarios covered by the model may have a leak point associated with well systems,
process systems or utility systems, and are described in Table 3.1. All leak scenarios normally

considered in a QRA (as defined in Chapter 2) and the scenarios covered by the model are
visualized in Figure 3.1.
Incidents occurring during well interventions/operations, such as wire line and coiled tubing, are
defined as blowouts or well releases, and are covered by Ref. /4/ that is based on the SINTEF
Offshore Blowout Database. These incidents are not covered by the model.
The definition of well releases from producing well (with or without gas lift) and injection well
during normal production ensures that all recorded topside well releases occurring during normal
operation is included in risk assessments based on this model (PLOFAM). Review of recorded
leaks from wells during normal production in HCRD and in the SINTEF Offshore Blowout Data-
base has led to the conclusion that the frequency for such incidents given in Ref. /4/, which is
based on the SINTEF Offshore Blowout Database, is inaccurate. There are recorded leaks in HCRD
that is not found in the SINTEF Offshore Blowout Database. Hence, the frequency for well release
during normal production in Ref /4/ should not be included in risk assessments based on PLOFAM,
as this is already included.
The design specification of a vent is exceeded when the released amount generate a fluid
concentration that is beyond the hazardous zone classification surrounding the vent. Such
incidents are covered by the model. See also Chapter 5.3 for considerations regarding safety
design of vents, and how risk associated with vent leaks should be managed.
Table 3.1 - Leak scenarios covered by the model. They occur in well system, process system or utility
system (process leaks fed through utility systems). Scenarios that are not listed in this table are not
covered by the model
Leak point in well system Leak point in process Leak point in utility system
system
1. Producing well/Injection 4. Leak point in pro- 5. Leak point in flare system (low
well: Topside well release cess system pressure or high pressure flare
where the inventory bet- between PWV and system)
ween DHSV and PWV is topside riser ESDV/- 6. Excessive releases through flare
released during normal storage ESDV. The tips and atmospheric vents that
production. fuel system is exceed the design specification
2. Gas lift well: Topside well regarded as part of and pose a fire and explosion
release where the inven- the process system. hazard to equipment, structures
tory between the ASV or personnel. Such leaks are de-
and the barrier towards noted vent leaks (see also
the process system is Chapter 5.3).
released. In cases where 7. Leak point in utility systems that
no ASV is present, the is fed by hydrocarbons stemm-
entire inventory in the gas ing from process system.
lift annulus to the ASCV Systems covered by the model
may be released. Assu- are:
ming that the check valve
a. Open drain system
ASCV is functioning,
otherwise there is no b. Closed drain system
barrier towards the reser- c. Chemical injection systems
voir.
3. Release of hydrocarbon
fluid from annuli that are
not used for gas lift

Figure 3.1 - Illustration of leak scenarios normally considered in a QRA (Ref. Chapter 2). The figure shows which scenarios that are covered by the
model and which that are not

4 Leak scenarios modelled in QRAs
The description of leak scenarios covered by the model, described in Chapter 3, gives guidelines
for selection of relevant historical incidents in the data used as basis for estimation of leak fre-
quencies and hole size distributions for process leaks and well releases during normal operation.
In other words, historical incidents that fall within one of the leak scenarios given in Chapter 3,
form the basis for frequency estimation and estimation of hole size distributions for the leak
scenarios modelled in a QRA. The leak frequency model has to define leak scenarios relevant for
QRAs that captures the important risk contributing leak scenarios (given in Chapter 3), and where
leak frequencies can be estimated based on available data material. This chapter defines leak
scenarios suggested for leak modelling in QRAs performed for oil and gas facilities at NCS. The
leak frequency model gives leak frequencies and hole size distributions for the leak scenarios
defined in this chapter (see TN-6), which covers the leak scenarios described in Chapter 3.
Three main leak scenarios for modelling in QRAs are defined for the leak frequency model. That
is Process leak, Producing well leak and Gas lift well leak. Further details and recommendations
with respect to modelling in QRAs are given for these scenarios in Chapter 4.1, 4.2 and 4.3,
respectively.
0.1 kg/s is recommended as leak rate threshold for estimation of leak duration (both in terms of
calculation of fluid dispersion and fire duration) in a QRA, for all leak scenarios in open areas and
3
leaks in enclosures having a net volume more than 1,000 m and with ventilation rate of 12 ach
or higher. This is explained in detail in Chapter 4.4.
Chapter 4.5 describes the rational for the lower boundary with regard to aggregated released
amount of hydrocarbons (10 kg). The model distinguishes on leak scenarios where the total
released amount of hydrocarbons is 10 kg, and >10 kg. These leaks are classified as Marginal
leaks and Significant leaks, respectively.
In total six leak scenarios for modelling in QRAs are defined, that are summarized in Table 4.1,
and Figure 4.1. The table also shows which leak scenarios in Table 3.1 that are put as basis for
estimation of leak frequencies and hole size distributions for the six defined leak scenarios for the
model. In a QRA the risk, in terms of fire- and explosion load exposure to vulnerable equipment
and structures such as safety systems, pressurized equipment, load carrying structures and main
safety functions, associated with Marginal leaks can be neglected. However, the risk to personnel
associated with Marginal leaks should not be neglected. This is further described in Chapter 4.5.
The model does not give separate leak frequencies for scenarios where the initial pressure is
different from the operational pressure. The initial pressure must be assessed as part of the QRA,
but normally the initial pressure is assumed equal to the operational pressure. This is further
discussed in Chapter 5.1. Furthermore the model does not give generic frequencies for limited
leaks (as defined in Ref. /8/), which is discussed in Chapter 5.2. Finally, even if recorded incidents
of releases through vents that represented a potential major accident hazard have been included
as part of the validation of the model to account for the underlying generic risk associated with
such events (see TN-3), this does not mean that the model accounts for the risk associated with
improper safety design or operation of such systems for a specific installation. This is further
explained in Chapter 5.3.
Release of hydrocarbon fluid from annuli that are not used for gas lift (Scenario 3 in Table 3.1), is
not suggested included as a separate scenario for QRAs. Instead these scenarios are included as
part of the frequency for producing well leaks and gas lift leaks.

Table 4.1 - Defined leak scenarios for QRAs
Modelled leak scenario Leak scenarios included

Significant Scenario 4-7 in Table 3.1, released quantity >10 kg
Process leak
Marginal Scenario 4-7 in Table 3.1, released quantity 10 kg
1
Producing well leak Significant Scenario 1 and 3 in Table 3.1, released quantity >10 kg
Gas lift well leak Significant Scenario 2 and 3 in Table 3.1, released quantity >10 kg
Figure 4.1 - Illustration and summary of the leak scenarios to be modelled in a QRA
4.1 Process leaks

As defined in Table 4.1, Process leaks cover leak scenario 4-7 in Table 3.1. Note that the leak
frequency for process leaks estimated by the model does also account for leaks occurring in the
utility system, but being fed from the process system. This is done by including process leaks fed
through utility systems, but not equipment counts from utility systems as basis for the model
validation. This implies that utility equipment should not be counted as basis for estimation of
process leak frequencies. Furthermore the model does not give separate leak frequencies for
process releases through utility systems and through process system. This means that the QRA,
based on the frequency model, will not reflect the potential location of the leak sources, and that
the leak frequency contribution from utility systems will scale with the number of equipment
counts for process system.
1
The frequency for producing wells and injection wells are assumed to be identical. The leak scenario is
denoted producing well only

Significant process leaks should be modelled as full pressure leaks, i.e. normal operational
conditions should be assumed. The released amount is restricted by the emergency shutdown
system (ESD system) and blow down system (BD system) if such systems are in place and
functioning as intended. The time to initiation of these systems are to be defined in the QRA, and
based on that the leak duration and released quantity should be estimated as part of the QRA.
4.2 Producing well leaks

As defined in Table 4.1, Producing well leaks cover leak scenario 1 and 3, in Table 3.1. A
significant producing well leak should be modelled as a full pressure topside well release of the
inventory between the DHSV and PWV at normal operational conditions. The released amount is
restricted and controlled by the DHSV towards the well (which is closed from the onset of the
leak). On the downstream side the segment will be controlled by the PWV. The time to closure of
the DHSV and PWV is to be defined in the QRA.
Producing well leaks and gas injection well leaks are modelled equally, and the leak frequencies
for these wells are judged to be identical.
4.3 Gas lift well leaks

As defined in Table 4.1, Gas lift well leaks cover leak scenario 2 and 3 in Table 3.1. A Significant
gas lift well leak should be modelled as a full pressure topside well release where the inventory
between the ASV and the barrier towards the process system is released. In cases where no ASV
is present, or the ASV is not functioning, the entire inventory in the gas lift annulus may be
released. Thus, the released amount is restricted and controlled by the ASV towards the well if
present/functioning. Otherwise the check valve ASCV is the only barrier towards the well, if
functioning. If the check valve is not functioning and the ASV is not present/functioning there are
no barriers towards the well. On the upstream side the segment may be controlled by the
Annulus Wing Valve (AWV), Annulus master valve or topside ESV. In addition, it could also be
controlled by the emergency shutdown system (ESD system) and blow down system (BD system)
if such systems are in place, they are functioning as intended, and are connected to the gas lift
system. The time to initiation of all these systems is to be defined in the QRA.
4.4 Initial leak rate boundary for leaks considered in a QRA

The lower initial leak rate boundary suggested for quantitative risk analysis that is used as basis
for the definition of Marginal leaks (see Chapter 4.5) is set to 0.1 kg/s for typical offshore process
modules. The initial leak rate is the leak rate at the onset of the leak (t = 0). This is in accordance
with industry practice in Norway (Ref. /1/). The threshold is in accordance with the threshold for
reporting of leaks to the Petroleum Safety Authority in Norway.
The associated limitation with regard to geometry is stated below, i.e. the validity of the term
typical offshore process module.
The threshold of 0.1 kg/s is based on the following:
The generated volume of the flammable cloud from a stationary 0.1 kg/s leak is less than 1
3
m (calculated in Phast) except cases where the natural ventilation is poor and the module
3
volume is small. A flammable volume of 1 m implies an insignificant probability for delayed
ignition caused by exposure to objects intended for use in explosive atmospheres. The basic
3
ignition probability per m free flow volume is according to the MISOF ignition model
-6 3
(Ref. /5/ ) 8.010 per m exposed to flammable gas for 5 minutes

3
The generated explosion loads in case of ignition of a flammable volume of 1 m is
considered negligible except in marginal enclosures (see below)

The flame length generated from a 0.1 kg/s leak is about 5 meter. The probability for ex-
posure to neighbouring equipment causing rupture of the exposed equipment is considered
to be significant, but small. 0.1 kg/s is defined as the lower cut off in Ref. /6/. The probability
for immediate fatal exposure to personnel is considered to be small for such fires
The probability for immediate ignition of a 0.1 kg/s release is small. According to the MISOF
ignition model (Ref. /5/), the immediate ignition probability is 0.00015 for all types of
equipment types. For pumps, an additional ignition probability of 0.007 applies.
Leaks less than 0.1 kg/s may in some cases constitute a significant risk, and should be considered
in such cases. An example of such a situation is enclosures, where leaks less than 0.1 kg/s is likely
to constitute a significant risk contribution (Ref. /7/). This applies in particular to explosions, but
also the consequences from fires may become severe in such cases, in particular with regard to
exposure to personnel as the entire enclosure may be exposed to intolerable fire loads (which will
depend on the available air for the combustion process). Explosions in such enclosures may give
fragments resulting from disintegration of the enclosure itself causing escalation and fatalities to
personnel in the vicinity of the enclosure. In addition, a hydrocarbon fluid concentration close to
the flammability range may be fatal due to asphyxiation. Such concentrations may arise quickly in
small enclosures. Lastly, the leak statistics shows that operational causes are dominant, which
means that in many cases personnel are present at the scene of the leak, for instance occurring
due to improper operation. The combined effect of personnel being present inside the enclosure
and the quickly arising accidental loads thus should be considered for incidents enclosures.
Based on the above discussion, the lower initial leak rate boundary for small enclosures should be
evaluated specifically in each case.
3
The 0.1 kg/s boundary applies for leaks in enclosures having a net volume more than 1,000 m
and with ventilation rate of 12 ach or higher. The volume in this context is the free flow volume
within the boundary of the area being studied (e.g. walls, ceiling and floor). For enclosure volu-
3
mes less than 1,000 m and/or smaller natural ventilation rates than 12 ach, initial leak rates less
than 0.1 kg/s may generate a significant flammable cloud. The housing around compressors for
noise protection is one example of a small enclosure that may have a net volume less than
3
1,000 m .
The relation between enclosure volumes, leak rate, forced ventilation rate and resulting average
steady-state gas concentration inside an enclosure is shown Figure 4.2. Figure 4.3 gives the
resulting total mas of gas in the enclosure. The calculation of the relations is based on the
following properties of the natural gas:

3
Density: 0.76 kg/m

3 3
Lower Flammability Limit: 4 vol% (= 0.03 kg fuel per m given a fuel density of 0.76 kg/m ).
The leak frequency model can be used to estimate the frequency for leaks having an initial leak
rate less than 0.1 kg/s, but the results should be interpreted in light of the fact that model vali-
dation has been focusing on leaks having an initial leak rate greater than 0.1 kg/s. Special
considerations may therefore have to be performed in each case.
The maximum initial leak rate should be assessed based on the frequency of full rupture and
properties of the system being studied.

Figure 4.2 - Average steady-state gas concentration in enclosure as a function of enclosure volume
for three different constant leak rates
Figure 4.3 - Mass gas in module at LFL at steady-state condition in enclosure as a function of enclo-
sure volume, given for three different leak rates. In addition the mass gas in enclosure if the gas
concentration= LFL, is given as a function of enclosure volume

4.5 Significant leaks vs. Marginal leaks
As described above, Marginal leaks are defined as leak scenarios where the total released amount
of hydrocarbons is 10 kg, while Significant leaks are defined as leak scenarios where the total
released amount of hydrocarbons is >10 kg. This is based on the lower boundary with regard to
aggregated released amount of hydrocarbons, which is set to 10 kg for typical offshore process
modules. The boundary applies with regard to calculation of risk in terms of exposure to
vulnerable equipment and structures such as safety systems, pressurized equipment, load carrying
structures and main safety functions. Marginal leaks will not impair these safety functions/objects,
and do only pose a risk to personnel.
The threshold of 10 kg follows from:
The longest duration of a marginal leak down to the lower initial leak rate boundary
(0.1 kg/s, see Chapter 4.4) becomes 100 seconds (10 kg divided by 0.1 kg/s), which only will
in very unfavourable conditions have sufficient duration to cause rupture of pressurized
equipment and/or impair to safety critical equipment. Furthermore, the probability for
exposure to vulnerable equipment in such a scenario is considered remote
The largest possible combustible cloud resulting from an instantaneous release of 10 kg will
generate marginal explosion loads. Assuming that 50 % of the amount (5 kg) is forming a
cloud having a homogenous concentration somewhat above than the lower flammability
3 3
limit (0.05 kg/m ) gives a gas cloud with a volume of 100 m (which is considered to be an
upper estimate). The resulting combustible fluid for such releases is expected to in the range
3 3
10 100 m . A cloud having a volume of 10-100 m will in typical offshore process modules
generate marginal explosion loads. However, for small enclosures this should be investigated
specifically according to below

3
A flammable volume of 100 m (or significantly less in most cases) having a short duration
implies an insignificant probability for ignition due to exposure to objects intended for use in
explosive atmospheres (that potentially possess a failure mode that cause ignition upon
3
exposure). The basic ignition probability per m free flow volume is according to the MISOF
-6
ignition model (Ref. /1/) 8.010 given 300 seconds exposure time
An instantaneous release of 10 kg may expose internal escape ways inside the area where
the leak occur and in some cases also external evacuation routes to the evacuation means.
The exposure time of intolerable fire loads would be very short, and will in practice have no
significance for the actual performance of the evacuation ways to mustering area/life boats.
The threshold does not apply for exposure to personnel. The leak statistics shows that operational
causes are prominent. In some of these cases, personnel were present at the scene of the leak,
which should be reflected in the consequence model.
According to available leak statistics, the number of leaks where the released amount is less than
10 kg is significant. Examples of such scenarios are:
A leak caused by erroneous opening of a valve that is closed immediately upon detection by
the operator of the valve (either at location or remotely by operator in control room)
A leak point arising in an isolated segment of the process system (e.g. unintentional leak of
small amount of pressurized gas inside valve house when valve drain plug is opened in initial
stage of valve maintenance operation).
The causes for marginal leaks could be both technical and related to activity (e.g. maintenance),
but it is judged that activity is a dominant factor, and probably greater than for significant leaks.
Hence, it is reasonable to argue that the actual underlying frequency for marginal leaks will be
specific to platform specific organisational factors. The estimate for marginal leaks provided by
the model is generic and cannot be adjusted according to installation specific properties. This
means that the frequency for marginal leaks should not be broken down into sub categories of
incidents in terms of causes or successful vs. non-successful operator intervention. The estimate is
to be considered generic, and should only be combined with an analysis of the ignition control
barrier (i.e. to set the ignition probability) and the consequences to personnel.

In order to estimate the fire frequency, it is recommended to use the stated probabilities for
immediate ignition in the MISOF ignition model (i.e. Pim and Pim,pump). This does not mean that the
analysis does not have to confirm that delayed ignition can be neglected.
As for the lower initial leak boundary, the boundary of 10 kg for marginal leaks does not apply
3
for enclosures having a net volume less than 1,000 m . Further description of the basis for this
limitation can be found in Chapter 4.4. Small enclosures are common at installations at NCS. Risk
3
related to explosions in such enclosures (i.e. < 1000 m ) should be carefully evaluated, as ignition
probability is high (ref MISOF /5/), and the required leak rate and/or total leaked mass to form an
ignitable cloud in small enclosures are small. Explosions in such enclosures may give fragments
causing escalations and fatalities to personnel in the vicinity.
5 Leak scenarios not covered by the model

Zero pressure leaks and limited leaks are leak scenarios that have been frequently used in the
industry, as process leaks were divided into Full pressure leaks, Limited leaks and Zero pressure
leaks in the previous model, Ref. /8/. This is illustrated in Figure 5.1. Significant process leaks as
defined in this model are modelled similar as Full pressure leaks in the previous model (see
Chapter 4.1). Limited leaks and zero pressure leaks are not included in PLOFAM. The reason is
explained in Chapter 5.1 and 5.2 below.
Considerations regarding safety design of vents and how risk associated with vent leaks should
be handled in a QRA are given in Chapter 5.3.
Figure 5.1 - Illustration and summary of the leak scenarios to be modelled in a QRA, together with
the leak scenarios Full pressure leaks, Limited leaks and Zero pressure leaks defined in the previous
model. Previous model refers to SHLFM Ref. /8/

5.1 Zero pressure leaks
Zero pressure leaks are in Ref. /8/ defined as scenarios where the overpressure is virtually zero
(0.01 barg or less). This may be because the equipment has a normal operating overpressure of
zero (e.g. open drains), or because the equipment has been depressurised for maintenance.
Figure 5.2 and Figure 5.3, gives the fraction and number of leaks that were classified as Limited
leaks, ESD isolated leaks, Late isolated leaks and Zero pressure leaks in Ref. /8/, that in PLOFAM is
classified as Marginal leaks (<10 kg) and Significant leaks (>10 kg). Only leaks that were included
both in Ref. /8/ and in this analysis are included. 4.4 % were classified as Zero pressure leaks,
where 2.6 % of these are in PLOFAM classified as Marginal leaks. Due to the small fraction of
Zero pressure leaks with a released quantity above 10 kg, and the small risk contribution from
zero pressure leaks, zero pressure leaks are in PLOFAM included as process leaks, i.e. handled as
process leaks with back pressure similar to the operating pressure. Hence the model does not
give separate leak frequencies and hole size distributions for zero pressure leaks, as defined in
Ref. /8/.
Figure 5.2 - Fraction of leaks that were classified as Limited leaks, ESD isolated leaks. Late isolated
leaks and Zero pressure leaks in Ref. /8/, that in PLOFAM is classified as Marginal leaks (<10 kg) and
Significant leaks (>10 kg)

Figure 5.3 - The number of leaks that were classified as Limited leaks, ESD isolated leaks. Late
isolated leaks and Zero pressure leaks in Ref. /8/, that in PLOFAM is classified as Marginal leaks (<10
kg) and Significant leaks (>10 kg)
5.2 Limited leaks

Limited leaks are in Ref. /8/ included to take into account scenarios that have a reduced released
quantity and/or a reduced duration (compared to process leaks as defined in Chapter 4.1) due to:
2. Restrictions in the flowline.
3. Operator intervention (valve closed by operator intervention)
4. The leak is at a favourable location in terms for pressure conditions. This is relevant for liquid
releases if, for instance, the leak point is above the lowest point in the segment, or the
amount of liquid is limited by high points in the segments.
In addition it is reasonable to believe that many of the limited leaks have occurred in connection
with human operations. In these cases the released quantity may be significantly reduced (as
compared to a Process leak as defined in Chapter 4.1), as the volume and/or pressure is likely to
be less than at normal operation. In case of leaks caused by human operations, recorded inci-
dents have demonstrated that the leak is often terminated by the operator (for example if a valve
is opened erroneously, it can be closed by the operator within a short time). These leaks are likely
to be classified as marginal leaks using the suggested leak scenario definitions in this technical
note. Figure 5.4 gives the fraction of leaks that were classified as Limited leaks, ESD isolated
leaks, Late isolated leaks and Zero pressure leaks in Ref. /8/, that in PLOFAM is classified as
Marginal leaks (<10 kg) and Significant leaks (>10 kg). Figure 5.5 gives the fraction of leaks that
in PLOFAM is classified as Marginal leaks (<10 kg) and Significant leaks (>10 kg) that in Ref. /8/
were classified as Limited leaks, ESD isolated leaks, Late isolated leaks and Zero pressure leaks.
From Figure 5.4 it is seen that 62 % of the Limited leaks are in PLOFAM classified as Marginal
leaks. Thus the remaining 38 % that were classified as Limited leaks that are now modelled as
significant process leaks, were probably classified as limited leak due to one of the above reasons
(point 1-3 above). It is however not recommended to model limited leaks on a general basis in
QRAs based on this. Instead it is recommended that other physical restrictions in the process
system than the ESD and PSD valves, such as check valves, gravity (high leak points or high points
for oil leaks) and other flow restrictions should only be reflected if an explicit model of the system

is established. The properties of the restrictions including reliability, if relevant, must be described
as part of the analysis, and should also include human interventions.
The duration of limited leaks follows from the developed specific model of the process system
accounting for the actual flow restrictions.
Figure 5.4 - Fraction of leaks that were classified as Limited leaks, ESD isolated leaks, Late isolated
leaks and Zero pressure leaks in Ref. /8/, that in PLOFAM is classified as Marginal leaks (<10 kg) and
Significant leaks (>10 kg)
Figure 5.5 - Fraction of leaks that in PLOFAM is classified as Marginal leaks (<10 kg) and Significant
leaks (>10 kg) that in Ref. /8/ were classified as Limited leaks, ESD isolated leaks, Late isolated leaks
and Zero pressure leaks

5.3 Vent leaks within design specification
Excessive releases through flare tips and atmospheric vents that exceed the design specification
and pose a fire and explosion hazard to equipment, structures or personnel are included in the
model. However, vent leaks where the release is within the design specification are not included.
In general, the safety design of vents is to be verified through other safety studies than the QRA.
Such studies include Hazard Operability Studies and specific assessments of the dispersion of the
fluid being released.
The safety design of the vents should be raised in the hazard identification process in a QRA
where the leak frequency model is applied. If there are known design issues with vent systems,
specific assessments should be performed (for example reliability of systems preventing overfilling
diesel tanks). The result should be considered included in the QRA if no technical solution that
rectifies the issue is found and the risk is judged to be significant. A generic frequency for vent
leaks that represent a significant risk is however not meaningful in the context of a QRA, i.e. pro-
viding additional information on how to control the risk associated with such systems.
Recorded incidents of releases through vents that represented a potential major accident hazard
have been evaluated as part of the project (see TN-3). Such known incidents have been included
as part of the validation of the model to account for the underlying generic risk associated with
such events. However, this does not mean that the model accounts for the risk associated with
improper safety design or operation of such systems for a specific installation. It must be noted
that the uncertainty associated with logging of vent releases is judged to be prominent as it is
expected that such leaks tend to only be reported if they are detected automatically by exposure
to detectors. Furthermore, it may be difficult to filter out such events in the data basis. As opera-
tional failures tend to be a dominate cause for vent leak scenarios it is difficult to establish a vent
release frequency even if the number of releases should be known due to lack of reliable
exposure data, i.e. number of vents and tank operations.

6 References
/1/ Standard Norge (2010). Risk and emergency preparedness assessment. Edition 3,
October 2010. NORSOK standard Z-013. Standards Norway (www.standard.no).
/2/ ESRA seminar, September 3, 2014, Stavanger, Per Holand, ExproSoft, Introduksjon til
gasslftbrnner for risikoanalytikere (See: http://esra.no/wp-content/uploads/2015/-
04/2-Holand-Intro.-til-gasslftbrnner.pdf)
/3/ Norwegian oil and gas association: An introduction to well integrity, Rev. 0, Date: 4
December 2012 (see https://www.norskoljeoggass.no)
offshore blowout database 2014, 17 March 2015, Report No: 19101001-8/2015/R3
Rev: Final
/5/ Lloyds Register Consulting Norway. Modelling of ignition sources on offshore oil and
gas Facilities - MISOF, Report no: 102657/R1, Date: 18. September 2014.
/6/ Scandpower Risk Management AS" Guidelines for the Protection of Pressurized Systems
exposed to Fire", Report No. 19101002/R1, Version 2. Date:1 January 2011
/7/ Stian Hiseth, Statoil Hydro ASA, Ingar Fossan, Scandpower AS, yvind Kaasa, Statoil
Hydro ASA; Managing Explosion Risk in Arctic Areas, SPE 111583, 2008 SPE Interna-
tional Conference on Health, Safety and Environment in Oil and Gas Exploration and
Production held in Nice, France, 15-17 April 2008.
1768, rev. 1, Date. 16.01.2009.

Working together
for a safer world
Leak frequency model

TN-5
Technical note for:


Date: 18 March 2016
1 Introduction .....................................................................................................................................1
2 Model summary and application of the model .................................................................................1
2.1 Model summary ....................................................................................................................1
2.2 Application of the model .......................................................................................................3
3 Equipment types covered by the model ...........................................................................................3
4 Leak frequency explanatory variables ...............................................................................................4
5 Requirement to the model and basic assumptions ...........................................................................9
5.1 Requirement to the model .....................................................................................................9
5.2 Basic assumptions..................................................................................................................9
6 Rational for the mathematical function for hole size distributions ..................................................10
6.1 Equipment size independent hole size distributions .............................................................10
6.2 Equipment size dependent hole size distributions ................................................................12
7 Equipment diameter dependencies ................................................................................................16
8 Lower hole size applicable for the model .......................................................................................16
9 Model summary and complexity of the model ...............................................................................17
10 References .....................................................................................................................................20
Appendix A - Guideline for equipment counts used as basis for leak frequency estimation in QRAs
Appendix B - Equipment size dependent model with reduced probability of large hole sizes
Technical note no: 105586/TN-5 Rev: Final B Page ii

1 Introduction
This technical note (TN) describes the leak frequency model basis and framework according to
the following:
The explanatory variables put down as basis for the model
Requirements and specifications to the model
Assumptions that the model are based upon
The rational for the mathematical formulation describing the hole size distributions and a
description of how equipment size is captured by the model.
In the initial section, Chapter 2, the model is summarized. This includes guidelines for application
of the model and how equipment counts (based on P&IDs) used as input to the model should be
executed.
expressions.
2 Model summary and application of the model

This chapter summarizes the mathematical formulation of the model. Furthermore, it gives
guidelines for application of the model and for equipment counting used as basis for leak
frequency estimation in QRAs. The rational and detailed description of the model is given later in
this technical note.
2.1 Model summary

The most general formulation of the mathematical equations for the complementary cumulative
hole size frequency distribution , which for simplicity is referred to as the hole size frequency
distribution (see TN-1), is:
(1)
() () , 1 <
(, ) = 0
0 , >
0 () = ( = 1, ) = () = (0 0 ) (2)
() = ( = , ) = 0 () () = 0 () ( + ) (3)
(4)
log( ) log( + )
0
() = =
() log()
where the parameters in the equation are described in Chapter 6 and 7. A summary is also pre-
sented in the table below, while a list of parameter values for , 0 , 0 , , and nec-
essary to estimate leak frequencies for all equipment types are given in (TN-6).
Note that this formulation is the general formulation for all equipment types. For several equip-
ment types, many of the parameters are set to 0 or 1, resulting in a simpler formulation for that
particular equipment type.
Technical note no: 105586/TN-5 Rev: Final B Page 1

Table 2.1 - Summary of all parameters used for each equipment type in the model. Note that not
all parameters are included in the above equations. Some are introduced later in the technical
note
(, ) Hole size frequency distribution (see TN-1) [year-1 component-1].
0 Total leak frequency [year-1 component-1].

F0 = F(d = 1).

FD = F(d = D).
Slope parameter

-1 -1
The fraction of the relevant equipment type with size D


2.2 Application of the model
When applying the model on a specific installation the first step is to define the desired leak rate
intervals defined by the leak rates 1 < 2 < < . Next the following procedure should be
followed for all equipment types on the installation:
1. Calculate 0 () for the relevant equipment
4. Calculate the hole size intervals defined by 1 < 2 < < corresponding to the defined
leak rate intervals 1 < 2 < < based on appropriate equations for modelling of leak
rate. TN-6 describes how this is done in the validation process. The models used for
estimating release rates should be carefully chosen based on fluid composition and process
conditions (e.g. pressure, composition and temperature). Since the model is defined for hole
sizes >1 mm only, it is recommended to set to 1 mm if the calculated hole size is < 1mm.
Thus leak frequencies for hole sizes less than 1 mm is not included. This will in general not
affect results in most QRAs as leaks around 1 mm will produce small release rates (< 0.1
kg/s). In some cases, leaks having a release rate less than 0.1 kg/s ought to be assessed in the
QRA to model the risk picture with adequate precision (e.g. enclosures with poor ventilation).
A special assessment of leaks with an initial leak rate less than 0.1 kg/s has to be performed
in such cases. The HCR data of leaks having hole size 1 mm or less (see TN-3) can be used as
basis for such a study
5. Calculate leak frequencies for all hole sizes 1 , 2 , for the relevant equipment based on
the equations given in Chapter 2.1
6. Calculate the frequency ( < < +1 ) = ( ) (+1 ) for the hole size intervals and/or
leak rate intervals
7. Multiply the leak frequencies with the number of equipment for the relevant equipment
type. The number of equipment on an installation should be estimated based on equipment
counting on P&IDs or similar. The exception is hoses where the frequency is multiplied with
the number of hose operations. The number of hose operations must be clarified with the
operator of the installation. A guideline for counting of equipment used as basis for leak
frequency estimation is given in Appendix A.
3 Equipment types covered by the model

In total 20 different equipment types are covered by the model, including Gas lift well and
Production well, which belongs to the well system. The other equipment types included in the
model are the most common process equipment types at offshore installations. All equipment
types covered by the model are given in Table 3.1.
Table 3.1 - Equipment types included in the model

Air-cooled heat exchanger
Atmospheric vessel Vessels with atmospheric pressure
Centrifugal compressor
Centrifugal pump
Compact flange

Filter
Flexible pipe
Hose Temporary hoses
Instrument
Pig trap Pig launchers and pig receivers
Plate heat exchanger
Process vessel Pressurized process vessels
Reciprocating compressor
Reciprocating pump
Shell and tube side heat Includes both shell side heat exchangers and tube side heat
exchanger exchangers
Standard flange Includes all flange types, except compact flanges
Steel pipe Process steel pipe
Valve Includes all types of valves
Gas lift well Well head with gas lift
Producing well Well head with or without gas lift
4 Leak frequency explanatory variables

The previous leak frequency model, Ref. /1/, was based on the fundamental presumption that the
leak frequency is proportional to the number of equipment of each type.
Based on NCS population data established in the project (see TN-2) the recorded number of leaks
at installations at NCS can be plotted against the number of equipment years per installation. The
results are presented in Figure 4.1 and Figure 4.5, where leaks with initial release rate >0.1 kg/s
and 10 kg/s are shown, respectively. In these figures all equipment types are included.
2
The coefficient of determination, i.e. the R -value of the linear regression displayed in Figure 4.1
is 0.61 for leaks with initial leak rate >0.1 kg/s, while the sample correlation coefficient given by
the general formula
( ) ( ) (5)
=
( )2 ( )2
is 0.79, which indicates a strong correlation. In Figure 4.2 - Figure 4.4, only valves, standard
2
flanges and instruments are included, respectively. The coefficient of determination, i.e. the R -
value of the linear regression, is 0.35, 0.29 and 0.06, while the sample correlation coefficient is
0.59, 0.54 and 0.26, respectively. The correlation is stronger if all equipment types are considered
together, than if each is studied separately. Based on the figures below, the number of
equipment is concluded as one explanatory variable.
It should be noted that for leaks with initial release rate >10 kg/s, the correlation is not that
strong, indicating that there are other explanatory variables for the large leaks. This has also been
pointed out at an early stage in the project by Lundin based on experience with large leaks.
However, other explanatory variables for large leaks have not been investigated in the project.
Thus they are treated similar to smaller leaks in the model.

Figure 4.1 - Number of recorded leaks with initial leak rate >0.1 kg/s vs. recorded equipment years
at installations at NCS. Note that steel pipes are not included. Each dot represents one installation
Figure 4.2 - Number of recorded leaks from Valves with initial leak rate >0.1 kg/s vs. recorded valve
years at installations at NCS

Figure 4.3 - Number of recorded leaks from Standard flange with initial leak rate >0.1 kg/s vs.
recorded Standard flange years at installations at NCS
Figure 4.4 - Number of recorded leaks from Instruments with initial leak rate >0.1 kg/s vs. recorded
instrument years at installations at NCS

Figure 4.5 - Number of recorded leaks with initial leak rate >10 kg/s vs. recorded equipment years
at installations at NCS. Note that steel pipes are not included. Each dot represents one installation
To investigate if also work orders (WO) on HC-equipment could be used as an explanatory vari-
able, Safetec has performed dependence test using Pearson correlation test, and a distance corre-
lation test. Compared to the correlation test, the distance correlation test will not be similarly
sensitive to the fact that both the number of leaks and equipment years are positive defined
variables. The tests are performed by the use of leak frequency data from RNNP from the period
2001 2014 for 34 installations with the same operator, and WOs collected for the year 2013.
Based on the following BORA classifications:
Operational failure: BORA classification B and C
Non-operational failure: BORA classification A, D, E and F
the tests are performed for three types of leak frequency:
Total leak frequency
Operational failure leak frequency (BORA cat. B and C)
Non-operational failure leak frequency.
In this way it can be investigated if the operational and non-operational leak frequencies have
different types of relations with the explanatory variables. The test used a confidence interval of
1 % (p-value 0.01). The results are given in Table 4.1 below. It is seen that equipment is
seemingly a stronger explanatory variable than work orders, but both could be used as
explanatory variables. ). These results cannot be conclusive as the activity data is less detailed
than the equipment count. The equipment count, represented by the estimated leak frequency
using SHLFM, is based on 18 different equipment categories benchmarked with historical leak
frequency, while the activity data is based solely on WO (on HC equipment). In order to have a
valid comparison, leak frequencies should be calculated using the activity based model and these
calculated frequencies should be used as explanatory variables representing activity level (Activity
based model split activity in three groups; Extensive WO, Limited WO and normal operations). It
is considered likely that such an approach would show that the relevance of activity is
comparable with equipment.

Table 4.1 - Results from dependence test using Pearson correlation, and a correlation test for both
operational related leaks, non-operational leaks and all leaks. Light blue indicates that the p-value
is >1 %, while dark blue indicates that the p-value is <1 %, and thus concluded to be strong
explanatory variable
Work orders could be used to include activity in the leak frequency model. There are several
reasons why a leak frequency model where activity is taken into account would be beneficial:
Enhances understanding of risk drivers
Focus both on consequence of leaks and reducing probability of leak
Reflects segment specific issues (sand, corrosive fluid/gas etc.)
Takes into account that different operations has different leak potential
Is in accordance with management regulations 4
The model could easily be used to analyse high activity periods.
Despite the fact that approximately 50 % of all leaks at NCS are related to activity, and the above
indices that WO can be used as an explanatory variable, the project concluded not to implement
WO as an explanatory variable for the model. This is partly based on the following:
The number of WOs is correlated with the number of equipment
Data is not publicly available
And use and duration of work orders may differ severely between operators
It is considered too challenging to include a model reflecting activity in this project.
Thus the number of equipment (for each equipment type) is the only explanatory variable
assumed in the model.

5 Requirement to the model and basic assumptions
This chapter presents technical requirements to the model. Furthermore, the basic assumptions
that the methodology is built on is presented.
5.1 Requirement to the model

The following technical requirements are identified for the model:
1. The model should be able to reproduce the total number of leaks in the period 2001-2014,
observed for a representative subset of installations at NCS (the larger number of installations
included, the better)
2. The model should be able to reproduce the total cumulative leak rate frequency distribution
experienced on NCS when applied to a representative subset of installations at NCS (the
larger number of installations included, the better)
3. The model should be able to reproduce the relative distribution between equipment types
seen based on experienced data from NCS and/or UKCS
4. The model should be able to reproduce the relative distribution between oil and gas leaks
seen in the experienced data from NCS and UKCS
5. The model should be able to reproduce the relative leak rate distribution per equipment type
seen in the experienced data from NCS (and UKCS). Stochastic effects are expected to be
prominent in this regard as the number of incidents will be few
6. The hole size probability distributions for the different components should describe the
properties of the observed hole size distributions extracted from the HCR data base as closely
as possible
7. The model must account for uncertainty in the underlying data basis as well as the stochastic
effects related to the observed phenomena
8. The model should be equipment size dependent and valid for single components
9. The model should be as simple as possible. i.e. the simplest possible model taking account
for trends, and effects that are justified through available data or argumentation, should be
chosen
10. The model should be robust towards changes in the dataset.
5.2 Basic assumptions

The following basic assumptions are made:
1. Data from NCS are the most relevant data for validation of the model. Where data from NCS
is not sufficient, data from UKCS is regarded as the best available alternative. Data from
UKCS will be important to establish the properties of the cumulative hole size probability
distribution for the different types of components
2. A model that fulfils the requirements described in Chapter 5.1 is assumed to produce the
best estimate for future leak frequencies
3. Each equipment has an associated leak frequency with a continuous holes size distribution.
Both the frequency and the hole size distribution is dependent on the type of process
equipment
4. The subset of installations used for validation (62 installations) is assumed representative for
all installations at NCS
5. The calculated annual leak frequency is representative for the whole period considered for
validation (2001-2014). The actual number of leak sources will be varying throughout the
period according to executed modifications to the platform

6. The underlying hole size distribution is the same for gas and oil leaks. The validity of this
assumption will be investigated as part of the validation process.
6 Rational for the mathematical function for hole size

distributions
This chapter describes the rational and the framework for the mathematical formulation of the
model. Chapter 6.1 presents the rational for describing the hole size frequency distribution as a
power law and how an equipment size independent model can be build using a modified power
law with a hole size dependent slope parameter. In Chapter 6.2 an explanation why the modified
power law with a hole size dependent slope parameter gives a good fit to the data is presented.
It is further described how this understanding can be used to describe an equipment size depen-
dent model. The model presented in Chapter 6.2 is the final mathematical formulation of the
model.
During the project the mathematical formulation of the equipment size dependent model pre-
sented in Chapter 6.2, was further developed to have the possibility to take expected properties
of the hole size distribution into account, i.e. a parameter 1 was introduced to take account for
a reduced probability of hole sizes slightly smaller than the equipment diameter. The mathemati-
cal description of this model, and the rational for it, is presented in Appendix B. As it turned out
that the benefit of this model was small, while the added complexity was significant, the model
presented in Appendix B was not chosen as the preferred leak frequency model. In order to
document the work done in the project and to report potential useful material for later model
development, the more complex model therefore documented in Appendix B.
6.1 Equipment size independent hole size distributions

One of the model requirements is that the model should give a continuous hole size distribution
for each equipment (ref Chapter 2). HCRD incident data indicates that the hole size frequency
distributions can be modelled by a power law relation (see Figure 6.1):
() = 0 (6)
Where 0 = ( = 1 ) is the total leak frequency for holes >1 mm from the relevant equip-
ment type, is the hole size and is the slope parameter (further described below). This is supp-
orted by the NCS data which indicates that the complementary cumulative leak rate distribution
follows a power law.
It should be emphasised that by using hole sizes with dimensions, the dimensions of 0 becomes
-1
different from year . This could have been solved by replacing with , where 0 =1 mm.
0
However this would result in dimensionless x-axis in the figures below. As, this could cause
confusion, and be a source of error, it is chosen to stick to the equation above, and give the
dimension at the x-axis of all figures.

Figure 6.1 - Curve fitting to recorded HCRD hole size distribution for Standard flange (Medium
size: 3-11). The data shows a power law behaviour
Three basic strategies for building a model exist:

a) Build the model solely based of available data material. This will of course require good
quality of the data, but it will also require the format of the data to be on the desired format
and the necessary level of detail
b) Build the model based on rational arguments of expected properties of the model
c) A combination of a) and b).
As one of the requirements defined in Chapter 5.1 states that the model should be equipment
size dependent and valid for single components, strategy a) will not give a satisfactory model.
The reason is that the HCR-data does not contain information about the equipment size as part
of the incident data (in late project phase equipment dimensions were made available to the
project for valves, flanges and pipes, but not other equipment types. However, as the number of
different equipment sizes is high, the number of incidents for each equipment size would most
likely not be sufficient to build hole size distributions on). Thus, curve fitting to hole size distribu-
tions would represent hole size distributions for a range of equipment sizes, where the equip-
ment size distribution is unknown. This implies that the correct usage of such a model would be
to let this hole size distribution be valid for all equipment sizes included in the underlying data
material. Thus, such a model would not be able to distinguish on equipment size, as all equip-
ment would give frequency contribution for hole sizes up to the upper validity range of the
model, i.e. most equipment would give frequency contribution at hole sizes, d, larger than the
equipment size, D. Implicitly, the underlying assumption for such a model, when applied on a
specific installation, would be that the equipment size distribution at the installation is
comparable to the equipment size distribution in the underlying data material (HCRD).
As part of the model development the project did curve fit the above power law equation to the
recorded hole size distributions for every type of equipment. It turned out that a better fit was
achieved if the slope parameter was expressed as a function of the hole size d as follows:
= log() + (7)
Where a and b are constants (normally negative). This function has the ability to reduce the slope
parameter for large hole sizes as often seen in the recorded hole size distributions in HCRD.
Figure 6.2 illustrates this for centrifugal compressor. Note that a constant slope parameter, is

achieved by setting a=0. Thus, curve fitting the power law with = log() + , will give
better fit to the data than if m is assumed constant, if a 0.
Figure 6.2 - Curve fitting to recorded HCRD hole size distribution for centrifugal compressor. A
better fit is achieved using equation (7) for the slope parameter, m (thin blue curve), than by
assuming a constant slope parameter (thick blue curve), m. Existing model refers to the SHLFM,
Ref. /1/
6.2 Equipment size dependent hole size distributions

In order to develop an equipment size dependent model, one will have to require the following:
1. ( = 1) = 0
2. ( = ) =
3. ( > ) = 0
Where is the full bore hole frequency. The frequency values for d < D, will be decided by the
assumed mathematical function representing the hole size distribution. As described above, the
curve fitting to recorded hole size distributions has shown that a better fit is achieved if the slope
parameter , is expressed as a function of the hole size d, as described in equation (7). This can
be explained by the fact that the weighted sum of a range of power law hole size distributions
will fit to a power law with decreasing slope parameter with increasing hole sizes (as described in
equation (7), d, as illustrated in Figure 6.3 . The upper figure gives the complementary cumulative
equipment size distribution taken from NCS. Assuming that the hole size probability distribution
from equipment of the same dimension follows a power law, given by ()/0 , and putting the
equipment size distribution for valve as basis for the range of power law functions in the lower
figure, results in the sum given by the light green curve in the lower figure (it is also assumed that
the rupture fraction follows the equation in Chapter 7, with =2.65, =-1,25 and
=0.00066, which are the parameters for valves, see TN-6 ). The black dotted line is a power
law with a d-dependent slope parameter, m as described in equation (7) (a=-0.5, b=-0.4). This
illustrates the point that the sum of power law functions will not be a power law if the
equipment size distribution is as given in the upper figure. In that case a power law with d-
dependent slope parameter will fit the sum better.

Figure 6.3 - above: The complementary cumulative equipment size distribution at NCS for standard
flange, valve and instrument. The distribution for valve is put as basis for the bottom figure.
Below: Each power law curve, is the weighted probability hole size distribution for a specific
equipment size. The probability hole size distribution is given by ()/0 . The light green curve is
the sum of all power law curves. The weighted sum of power laws (light green curve) can be fitted
using the hole size dependent slope parameter given in equation (7). The black dotted line is
achieved by setting a=-0.5 and b=-0.4
There may be explanations why the mathematical function describing the equipment size depen-
dent hole size distribution should also have d-dependent slope parameters. This would for in-
stance be the case if each failure mode is associated with a unique power law, and the total hole
size distribution is a weighted sum of contributions from different failure modes. However, docu-
mentation of this, requires a level of details that is higher than available in this project, and in
order to keep the model as simple as possible, as required in Chapter 5.1, it is assumed that the
hole size frequency distribution follows a power law in the range d < D.

It follows from initial condition 1 and 2 above that the slope parameter is given by the
following expression:
(8)
log( ) log(0 ) log( )
0
= =
() ()
This is also illustrated in Figure 6.4. The complementary cumulative probability distributions, given
by ()/0 , and underlying probability density distributions are illustrated in Figure 6.5 for a
range of equipment sizes. The latter distribution is important as it is easier to relate actual
physical properties of the holes to the distribution, which in particular is useful for the analysis
presented in Appendix B. It gives the probability of holes within a hole size range . In the
below figure is 1 mm. This results in a spike with a value corresponding to the full bore hole
frequency . Note however that the probability density function value for d=D, ( = ),
depends on the size of . If 0, ( = ) .
Figure 6.4 - Illustration of the slope parameter, for a 4 inch equipment

Figure 6.5 - Above: Illustration of the hole size probability distribution, given by ()/0 , for a
range of equipment diameters (the legend gives the dimension in inches). Below: The underlying
hole size probability density distribution

7 Equipment diameter dependencies
Both the total frequency 0 , and the full bore hole frequency, , is in general assumed to be de-
pendent on the equipment diameter . The following general dependencies are assumed:
0 () = () = (0 0 ) (9)
() = 0 () () = 0 () ( + ) (10)
The parameter gives the average leak frequency (independent of equipment diameter) for
the relevant equipment type and can be estimated based on historical leak data. Thus it is requi-
red that
(11)
= 0 () = ()

where the parameter is the fraction of the relevant equipment type with size D, and could be
estimated based on available population data. This gives the following constraint for the para-
meters 0 and 0 :
(12)
1 = () = (0 0 )

The parameter () gives the adjustment factor of the leak frequency for the relevant
equipment diameter relative to the weighted average leak frequency .
()
() = , gives the fraction of the full bore hole frequency for the relevant equipment
0 ()
diameter to the total leak frequency for the relevant equipment diameter. By defining the
equipment size dependencies as described above gives the flexibility to estimate 0 () and
() as a constant ( = 0), a linear relationship ( =1) and a power law relationship (B = 0) and
is therefore a general formulation that is able to reproduce trends seen in the data material.
8 Lower hole size applicable for the model

The lower hole size that the model is valid for is set to 1 mm. This chapter presents a discussion
that justifies the lower hole size applicable for the model.
TN-6 presents the model validation, and shows that the model reproduces the number of
observed leaks > 0.1 kg/s. This is based on the NCS population dataset, which contains
equipment counts from 62 installations at NCS (see TN-2), and on equations for leak rate
estimation for gas and liquid leaks (see TN-6). Figure 8.1 gives the hole size that gives 0.1 kg/s
leak rate as function of gauge pressure for Methane, Ethane, Propane and oil with density 800
kg/m. The leak rate equations used in the validation model are applied. The NCS population
database shows that the majority of equipment is associated with a pressure in the range 10
150 barg. For gas leaks, the hole size generating a leak rate of 0.1 kg/s, which is the minimum
leak rate used in the validation process, is in the range 2-9 mm (see Figure 8.1). For liquid leaks
the corresponding range is 1-2 mm. Thus the model can be said to be validated for hole sizes
down to 2 mm for gas leaks and 1 mm for liquid leaks. This puts confidence behind the lower
hole size that the model is valid for, which is set to 1 mm.
Note that the model may also be applicable also for lower hole sizes, but this has not been
validated.

Figure 8.1 - Hole size that gives 0.1 kg/s leak rate as function of gauge pressure for Methane,
Ethane, Propane and oil with density 800 kg/m
9 Model summary and complexity of the model

Ultimately the hole size frequency distribution deduced by the model, is given by the
6 parameters , 0 , 0 , , and . It may seem like this gives a high degree of freedom
for the model, and an unnecessary high model complexity. This chapter is prepared to explain the
parameters, and why they are necessary to capture the most important effects relevant for the
model. Furthermore, this chapter compares the above set of parameters to the set of parameters
established in SHLFM, Ref /1/, , , , and , and relates the parameters to the explicit and
implicit model assumptions.
The model is built up of the following parts
1. Modelling of the total leak frequency per component, 0 . To model the equipment diameter
dependency of 0 , the parameters , 0 and 0 are used (See Chapter 7). Note however
that the constraint given by Equation (12), reduces the number of degrees of freedom. Note
also that the parameter has to be adjusted to fit historical data. Even if there is flexibility
with regard to relative contribution from the different components, the degree of freedom is
highly restricted by the constraint given by historical data (see TN-6)
2. Modelling of the full bore hole frequency . To model the equipment diameter dependency
of , the parameters , and are used (See Chapter 7)
3. Modelling of the frequency of hole sizes in the interval 1 mm < < . The model here
assumes that the hole size frequency distribution, follows a power law, that starts at
( = 1) = 0 and ends at ( = ) = .
Modelling of the steps 1-3 is regarded as a minimum to capture the most important effects of
the model. Table 9.1 gives a summary of involved model parameters necessary to model the
different parts of the model given above, both in PLOFAM and in SHLFM. Constraints reducing
the complexity and underlying assumptions are also given. The table shows that modelling of 0
requires three parameters in PLOFAM where a constraints reduces the degree of freedom to two.
In SHLFM the corresponding number of parameters is 4. To model , PLOFAM requires 6 para-
meters (taking the constraint into account this is reduced to 5), while SHLFM requires 5 para-
meters. In both cases all model parameters are needed to model . The increased complexity in

PLOFAM (compared to SHLFM) is necessary to be able to model the full bore hole frequency pro-
perly. Note that in SHLFM only one extra parameter is used to model as compared to 0 , while
PLOFAM adds 3 parameters, which is necessary to improve the modelling of large hole sizes,
which is one of the main reasons for updating the model. The slope parameter given by PLOFAM
is apparently complex, but this relation follows from the assumption of a power law relation for
leak frequency from hole sizes smaller than the equipment diameter (< ). No new parameters
are introduced. In SHLFM the slope parameter is assumed to be constant, which has implicit
implications for modelling of , as described in Table 9.1.
Table 9.1 - The table gives a summary of involved model parameters necessary to model the
different parts of the model given above. Constraints reducing the complexity and explicit and
implicit assumptions are also given
Part of Model parameters in new PLOFAM Model parameters in SHLFM
model
0 Involved parameters: , 0 and 0 Involved parameters:, , and
0
Assumption: 0 = (0 ) Assumption: 0 = (1 + )+
Constraint: 1 = (0 0 ). See
Chapter 7.
Involved parameters: , 0 , 0 , , Involved parameters: , , , and
, .
Constraint: Same as for 0 Assumption: = const
Assumption:
= 0 () ( + ) = The assumption that the slope para-
(0 0 ) ( + ) meter is constant (independent of
equipment size )
leads to the following implicit assum-
ption in SHLFM for :
= (1 + ) +
Involved parameters: , and . Involved parameters:
Assumption: ( = 1) = 0 , ( = ) = Assumption: = const
.
The slope parameter follows from the
assumption of power law relation for the
hole size frequency distribution:

log
0
() = =
log()
log( + )
log()
1 Not included in new model (see Appendix Involved parameters: 1 =

(denoted B for model alternative including 1 ) Assumption: 1 = =
in
SHLFM) :

The reason for the increased number of parameters in PLOFAM (compared to SHLFM) is the
improved modelling of full bore hole releases. This is regarded as necessary to improve modelling
of such releases, which is one of the most important features of the model.

10 References
1768, rev. 1, 16.01.2009.

Appendix A
Guideline for equipment

counts used as basis for leak
frequency estimation in QRAs

1 Introduction .................................................................................................................................. A1
2 Guideline for equipment counts used as basis for leak sources ..................................................... A1
2.1 Compressor ........................................................................................................................ A3
2.2 Flange ................................................................................................................................ A3
2.3 Filter ................................................................................................................................... A3
2.4 Heat exchanger .................................................................................................................. A3
2.5 Hose ................................................................................................................................... A4
2.6 Instrument.......................................................................................................................... A4
2.7 Pig trap .............................................................................................................................. A4
2.8 Process vessel ..................................................................................................................... A4
2.9 Pump ................................................................................................................................. A5
2.10 Steel pipe ........................................................................................................................... A5
2.11 Valve .................................................................................................................................. A5
2.12 Well ................................................................................................................................... A5

1 Introduction
This Appendix gives guidelines for equipment counting on P&IDs for use as basis for leak
frequency estimation using the leak frequency model.
2 Guideline for equipment counts used as basis for leak

sources
P&IDs (Process and Instrument Drawings) are normally used as basis for the counting of
components. Normally, a P&ID Legend is available for correct interpretation of the symbols used
on the drawings. P&IDs do in many cases not display all actual leak sources within a system, but
is in general accepted as the basis for equipment counts in QRAs. In particular flanges associated
with valves with small dimensions tend not to be included in P&IDs.
If more detailed information is needed, one can count from ISO-drawings or even from CAD-
models. In some cases, lists of components can be generated from other systems holding
information about equipment.
For calculation of leak frequencies for use in risk analysis, all leak sources in the process system
should be identified, as well as the number of well heads. Note that leak points in the utility
system should not be counted (ref TN-4). For system boundaries it is referred to TN-4. All
components that normally pressurized by the process system should be counted. In QRAs based
on PLOFAM, no adjustment of calculated leak frequencies should be performed to take limited
operational time into account (see TN-6, Chapter 11.4). Typical leak sources (components) in a
process plant are:
Compressor (Reciprocating and centrifugal)
Flange (Standard flange and compact flange)
Filter
Heat exchanger (shell/tube side heat exchangers, plate heat exchangers and air cooled heat
exchangers)
Hose
Instrument
Pig trap
Process vessel
Pump (Reciprocating and centrifugal)
Steel pipe (meter piping)
Valve
Well heads (Producing well head with gas lift, producing well head without gas lift, gas
injection well).
Guidelines for counting of these equipment types are given in the following sub-chapters. It is
also referred to leak incident registration rules given in HCR definitions (see TN-3 Appendix A).
Figure 2.1, gives an example of counting on P&IDs that is used as reference in several sub
chapters below.

Figure 2.1 - Example of counting on P&ID

2.1 Compressor
The following counting rules apply:
It is differentiated between reciprocating and centrifugal compressors in the leak model
The compressor is defined to have the same dimension as the dimension of the in/outlet
piping
Flanges connected to a compressor are counted as flanges
Only the equipment itself should be counted. i.e. the equipment should be counted
excluding all valves, piping, flanges, instruments and fittings
Each compressor on the P&ID is counted. Also if several compressors are driven by the same
shaft, they should all be counted as separately (for example if two compressors are driven by
the same shaft, they should be counted as two compressors).
2.2 Flange
Two flanges that are connected as one mechanical coupling (flanged joint) are counted as
one Standard flange or one Compact flange
One spectacle blind is counted as two flanges
Corrosion coupons are counted as flanges
An end flange on a pipe is counted as one flange
Inlets/outlets of a process package (e.g. a metering package, compressor, strainers, etc.) are
counted as flanges
The flange downstream of a closed valve shall not be counted. (This flange does not see any
pressure from the medium contained in the system)
The leak frequencies applied differentiate between a compact flange type (SPO flange) and a
standard ANSI or API flange type (standard offshore flange).
2.3 Filter
Only the equipment itself should be counted. I.e. the equipment should be counted excluding all
valves, piping, flanges, instruments and fittings.
2.4 Heat exchanger

The model differentiates between four types of exchangers:
o plate exchanger
o tube side heat exchanger
o shell side heat exchanger
o Air cooled heat exchanger
A heat exchanger is defined to have dimension as the dimension of the inlet/outlets.

2.5 Hose
Hoses are not counted in number of hoses, but in number of operations. Thus they are not
counted on P&IDs.
The number of hose operations must be clarified with the operator of the installation. The
number of hose operations at installations on NCS fluctuates significantly (see TN-2), and will
depend on the design solutions for equipment such as well heads and pig receivers. Therefore no
general guidance can be given for estimation of the number of hose operations. If the number of
hose operations at an installation is unknown, the described databasis for the model, described in
TN-2, can be used as reference.
Be aware that leaks from hoses constitute around 5% of all leaks at NCS in the period 2001
2014. The number of hose operations varies significantly among installations, due to differences
in both design and operational philosophies. This means that for some installations the
contribution from hoses will be considerable. The QRA should therefore base the leak frequency
from hoses on quality assured data for the installation in question.
See also TN-2 and TN-6 regarding uncertainties related to hose operations, and the way
uncertainties are taken into account in the model.
2.6 Instrument
All instrument are assumed to have dimension 0.75" (20 mm nominal diameter)
Instruments with two (or more) connection points to the process equipment are counted as
two (or more) instruments (e.g. level inductors on vessels). Examples of this are seen in
Figure 2.1
An instrument, including its valves and flanges, is counted as one instrument only. Hence,
these valves and flanges should not be counted separately. Examples of this are seen in
Figure 2.1
Only instruments that are connected to the process with valves or flanges are counted
(Instruments which are not connected in this way are located outside the process piping and
are therefore not leak sources).
2.7 Pig trap

Each item comprises the item of equipment itself, but excluding all valves, piping, flanges,
instruments and fittings beyond the first flange and excluding the first flange itself.
2.8 Process vessel

It is not differentiated between types of vessel in the leak statistics
The vessel should be counted as a vessel with size equal the main inlet/outlet of the vessel
Only the equipment itself should be counted. i.e. the equipment should be counted
excluding all valves, piping, flanges, instruments and fittings
Flanges connected to a vessel are counted as flanges
Man holes are regarded as part of the vessel and are not counted separately.

2.9 Pump
It is differentiated between reciprocating and centrifugal pumps in the leak model
The pump is defined to have the same dimension as the dimension of the in/outlet of the
pump
Flanges connected to a pump are counted as flanges
Only the equipment itself should be counted. I.e. the equipment should be counted
excluding all valves, piping, flanges, instruments and fittings.
2.10 Steel pipe

To estimate leak frequency from process piping, the number of meter of piping in the system in
question should be estimated and used as input to the leak frequency model for piping. If the
number of meter of piping is not available, the total contribution from piping is recommended
set to 14 %. Thus the leak frequency for all other equipment types should be multiplied by a
1
factor = 1.16 in order to get the total leak frequency including contribution from piping.
0.86
2.11 Valve
It is differentiated between ESVs and other valves in the model
A closed valve and an open valve are both counted as one valve
For info: P&IDs label valves as e.g. "LO" (Locked Open), "LC" (Locked Close), "KILO" (key
interlock open) or KILC (key interlock closed)
A valve can either be flanged or welded and P&IDs do not always show all flange connec-
tions related to valves. Hence, it is important to clarify, whether the valves are flanged or
welded. If the valves are flanged, two flanges are counted for each open valve, and one
flange is counted per closed valve (the flange upstream the closed valve). If the valve is
pressurized on both sides, both flanges are counted.
2.12 Well
The number of well heads should be counted. Other equipment such as flanges and valves on
the well head are not counted separately. The model distinguishes on:
Producing well head with gas lift
Producing well head without gas lift
Gas injection well head.
Well heads with gas lift should be counted both as a gas lift well and as a producing well (with-
out gas lift). If for example an installation has 15 wells, where 5 have gas lift, this should be
counted as 5 gas lift wells and 15 production wells.
The barriers between the process system and the well system are described in TN-4.

Appendix B
Equipment size dependent

model with reduced probability
of large hole sizes

1 Introduction ...................................................................................................................................B1
1.1 Model summary ..................................................................................................................B1
2 Equipment size dependent model with reduced probability of large hole sizes...............................B2
2.1 Model description and rational for the model ......................................................................B2
2.2 Implicit assumptions in the model........................................................................................B5
3 Model summary and complexity of the model ...............................................................................B6
4 References .....................................................................................................................................B9

1 Introduction
During the model development it was argued that there should be a very small probability of
occurrence of a hole that is slightly smaller than the equipment diameter. Therefore a model
where this effect was taken into account was established. However it turned out that the
benefits of this model was small, while the added complexity was significant. Therefore it was
decided to build the model as described in TN-5. However, in order to document the work done
in the project and to report potential useful material for later model development, the more
complex model is documented in this appendix. Readers familiar with SHLFM, Ref /1/, will notice
that the model documented in this appendix, has similar mathematical form as SHLFM, where a
constant ( ) was included.
1.1 Model summary

The most general formulation of the mathematical equations for the accumulated frequency hole
size distribution presented in this appendix is
[0 () ()] () + () , (1)
(, ) =
0 , >
0 () = ( = 1) = () = (0 0 ) (2)
() = ( = ) = 0 () () = 0 () ( + ) (3)
1 (4)
log( )
0 1
() =
()
where the parameters in the equation are described in Chapter 2 and in TN-5. A summary of
added model parameters not described in TN-5 are presented in the table below.
Note that the above formulation is the general formulation for all equipment types. For several
equipment types, many of the parameters could be set to 0 or 1, resulting in a simpler
formulation for that particular equipment type.
Table 1.1 - Summary of all parameters used in the model. Note that not all parameters are included
in the above equations, but are them introduced later in the technical note
-1 -1
1 Additional full bore hole frequency [year component ]
1
=


2 Equipment size dependent model with reduced
probability of large hole sizes
2.1 Model description and rational for the model
It can be argued that there should be a very small probability of occurrence of a hole that is
slightly smaller than the equipment diameter. Thus the power law dependency should be valid
only up to some critical hole size . In the region < < the probability should be very
small. It could potentially be zero and it could both be decreasing and constant. For = the
probability should correspond to the full bore hole probability, resulting in a spike in the
probability density distribution for d=D. This is illustrated in Figure 2.1, where the following two
functions are plotted to illustrate the concept:
1 = 1 ,
, (5)
(, ) = 2 = 2 ,
<
0 , >
1 = 1 , (6)
(, ) = = <
0 >

Figure 2.1 - Illustration of how the actual accumulated probability hole size distribution (above)
and the probability density hole size distribution (below) might look like. The blue curve is a
standard power law for a 4 equipment, while the red curves illustrate the two concepts described
in equation (5) and (6) above. F_m1, F_m2 and F_constant have identical line colour as they belong
to the same model.
To determine the exact hole size distributions and the parameter would require detailed
analysis together with specialists on every single equipment included in the model, and exceeds
the possible detail level in this project. However, a pragmatic solution to the challenge, that has
the ability to capture the described effects, would be to include an additional full bore hole
frequency 1 in the frequency equation.
() = (0 1 ) + 1 (7)
In order to keep 0 as the total leak frequency for the relevant equipment type, 1 must be
subtracted from the first factor in the equation. This function is shown in Figure 2.2, together
with the other possible model suggestions illustrated in Figure 2.1. The light green curve (1 ) and
the dark green curve ((0 1 ) ) are also plotted and adds up to the total frequency function
().

This solution has the advantage that it gives a continuous derivable frequency function, and that
it gives the same mathematical form as applied in the existing model, Ref /1/. Hence, it is familiar
to the industry. It should be noted that neither of the suggested models represent the correct
hole size distribution, but equation (7), has the capability to capture effects of the hole size
distributions that have been argued for in the project. The suggested model is assessed as the
simplest possible model that incorporates the effect of reduced probability of hole large hole
sizes (relative to the standard power law).
It follows that the slope parameter , can be expressed as
1 (8)
log( )
0 1
() =
()
The parameter 1 can be expressed as a function of as follows:
1 () = () , [0, 1 (9)
1
where the parameter = = is thus the fraction of the added full bore hole fre-

quency to the total full bore hole frequency () = ( = ). > 0 will give a curved line,
while = 0 will give a normal power law.
Figure 2.2 - Illustration of the suggested mathematical model F(d) (green curves) together with
other suggested models: Standard power law (blue curve) and a model where the probability of
hole sizes close to the equipment diameter is lowered compared to the standard power law (red
curve). Two different models for high hole sizes are illustrated: One second power law for dcrit < d
< D and one solution with constant probability for dcrit < d < D. In this illustration 0 =1e-4 per year

per equipment, and = 1e-6 per year per equipment, while = 1 = 0.85


2.2 Implicit assumptions in the model
This chapter describes the implicit assumptions in the model with respect to the parameter
defined in Chapter 2.1. This is presented as it is important to have an overview of the underlying
assumptions and behaviour of the model. In particular it is important to know how the model
relates to physical properties that are possible to make qualified assessment of, such as .
As described in Chapter 2.1, the mathematical model has the capability to capture the effect that
hole sizes close to the equipment diameter in the range < < is expected to be even
lower than estimated by the normal power law (the normal power law is achieved by setting =
0. By increasing , is reduced). The parameter can in the concluded model be defined as
the hole size where the hole size dependent part of the equation equals the full bore hole
frequency (this is in accordance with the definition in Chapter 2.1 if =0):

(0 1 ) = () = (0 1 ) + 1 (10)
It follows that can be expressed as follows:
1 (11)

= +
0
Figure 2.3 shows how varies with equipment diameter, D for different values of , and thus
the implicit assumption regarding when is kept constant (independent of D). Furthermore

the parameter may be defined as the fraction :

1 (12)

= = 1 +
(0 )
Figure 2.4 shows how varies with equipment diameter, D for different values of , and thus the
implicit assumption in the model regarding when is kept constant. The figure shows that for

low values of , the fraction is close to constant. For higher values of , the fraction decrea-

ses with increasing equipment diameter, D. The exact expected behaviour is difficult to state, but
it seems reasonable that decreases with increasing D. Based on this assessment it is concluded
that keeping constant gives a model that is in line with the projects understanding and assess-
ments of how hole sizes develop in equipment of different dimensions.

Figure 2.3 - as function of equipment diameter, D for different values of

Figure 2.4 - =

as function of equipment diameter, D, for different values of
3 Model summary and complexity of the model

TN-5 presents a summary of involved model parameters necessary to model the different parts of
the model (in the Chapter Model summary and complexity of the model), both for PLOFAM
and for SHLFM. This chapter gives explanations of the added model parameter (added in the
model described in this appendix) and model parameters influenced by . Including the
parameter, the model is described by the 7 parameters , 0 , 0 , , , and . The
parameters established in SHLFM, Ref. /1 /, are , , , and .
The model is built up of 4 parts, where parts 1-3 are described in TN-5. The additional model part
introduced in the model described in this appendix is:
4. To capture the effect that hole sizes close to the equipment diameter in the range <
< is expected to be even lower than estimated by the normal power law, the parameter
1 is introduced (see Chapter 2.1), which relates to through the parameter ( assumed
constant. The effect of a constant is analysed in Chapter 2.2).

Modelling of the steps 1-3 are regarded as a minimum to capture the most important effects of
the model, while step 4 has proven to give minor changes to the leak frequency distribution (see
TN-6), which is the reason why it is not included in the concluded model described in TN-5.
Table 3.1 gives a summary of involved model parameters necessary to describe the different parts
of the model given above both for PLOFAM and for SHLFM.
The slope parameter in PLOFAM is apparently complex, and introducing the parameter makes it
even more complex, but this relation given for the slope parameter follows from the assumption
of a power law relation for leak frequency from hole sizes 1 mm < < . No new parameters are
introduced. In SHLFM the slope parameter is assumed to be constant (independent of equipment
size ), which has implicit implications for modelling of , as described in Table 3.1. Finally,
introducing the parameter 1 , requires one extra parameter () to be defined in PLOFAM, while
SHLFM apparently does not need new parameters to be defined, as is already defined as a
part of 0 . This is however related to the way 0 is defined in the two models. Obviously the
parameter in SHLFM has the same effect as the parameter in PLOFAM. One can argue that
is not really necessary to model 0 and in SHLFM, and that being part of 0 and (in
table below) in SHLFM is related to the definition of 0 and in PLOFAM. In that case the
parameters necessary to model 0 and in SHLFM should be reduced by one, and the number
of parameters necessary to model 1 would then be one ( ).
Table 3.1 - The table gives a summary of involved model parameters necessary to model the
different parts of the model given above. Constraints reducing the complexity and explicit and
implicit assumptions are also given
Part of Model parameters in PLOFAM Model parameters in SHLFM
model
0 See TN-5 See TN-5
See TN-5 See TN-5
Involved parameters: , , and . Involved parameters:
Assumption: ( = 1) = 0 , ( = ) = Assumption: = const
.
The slope parameter follows from the
assumption of power law relation for the
accumulated leak frequency distribution:
1
log
0 1
() = =
log()
( + )(1 )
log
1 ( + )
log()
Note that setting =0 simplifies the
expression.
1 Involved parameters: , 0 , 0 , , Involved parameters: 1 =
(denoted , and . Assumption: 1 = =
in Constraint: Same as for 0
SHLFM) : Assumption: The assumption that is constant
1 () = () , [0, 1 (independent of equipment size )
= leads to the following implicit ass-
umption in SHLFM for :

Part of Model parameters in PLOFAM Model parameters in SHLFM
model
The assumption that is constant (inde- 1
() = =
pendent of equipment size ) ( + ) +
leads to the following implicit assumption
in PLOFAM for 1 :
1 () = [0 () ( + )] =
[ (0 0 ) ( + )]

4 References
1768, rev. 1, 16.01.2009.

Working together
for a safer world
Parametrization of model parameters

and validation
TN-6
Technical note for:


Date: 18 March 2016
1 Introduction .....................................................................................................................................1
2 General methodology description ....................................................................................................1
3 Model parameters ...........................................................................................................................3
4 Model targets ..................................................................................................................................4
5 Parameterization only based on UKCS data .....................................................................................4
5.1 Description of methodology ..................................................................................................4
5.2 Estimation of total frequency trend with equipment dimension .............................................5
5.2.1 Methodology...................................................................................................................... 5
5.2.2 Valve and instruments ........................................................................................................ 6
5.2.3 Standard flange .................................................................................................................. 7
5.2.4 Steel pipe ........................................................................................................................... 9
5.3 Full bore hole fraction .........................................................................................................10
5.3.1 Methodology.................................................................................................................... 10
5.3.2 Valves .............................................................................................................................. 13
5.3.3 Standard flange ................................................................................................................ 16
5.3.4 Steel pipe ......................................................................................................................... 17
5.4 Uncertainty small hole sizes .................................................................................................18
5.5 Fitting to observed frequency hole size distribution from HCRD ...........................................19
5.6 Shape of the hole size frequency distribution ......................................................................28
5.7 Modelling of various valve types ..........................................................................................29
6 Description validation model ..........................................................................................................33
6.1 Methodology.......................................................................................................................33
6.2 Release modelling................................................................................................................34
6.2.1 General ............................................................................................................................ 34
6.2.2 Gas releases ..................................................................................................................... 34
6.2.3 Liquid releases .................................................................................................................. 36
7 Validation of model based on UKCS data ......................................................................................37
8 Parameterisation only based on NCS data ......................................................................................53
8.1 General ...............................................................................................................................53
8.2 Hoses ..................................................................................................................................53
8.3 Steel pipe ............................................................................................................................55
8.4 Compact flange...................................................................................................................55
9 Best model fit to observed leaks at installations on the NCS...........................................................58
9.1 General ...............................................................................................................................58
9.2 Model parameters ...............................................................................................................58
9.2.1 Hoses ............................................................................................................................... 58
9.2.2 Producing well and Gas lift well ........................................................................................ 58
9.2.3 Contribution steel pipe ..................................................................................................... 59
9.2.4 Fraction marginal leak ...................................................................................................... 59
Technical note no: 105586/TN-6 Rev: Final B Page ii

9.2.5 Model for instruments, valves and standard flanges .......................................................... 59
9.2.6 Model for equipment with low number of incidents.......................................................... 60
9.3 Validation of the NCS model ...............................................................................................60
10 PLOFAM ........................................................................................................................................74
10.1 General ...............................................................................................................................74
10.2 Model parameters ...............................................................................................................74
10.2.1 Hoses ............................................................................................................................... 74
10.2.2 Producing well and Gas lift well ........................................................................................ 74
10.2.3 Contribution steel pipe ..................................................................................................... 74
10.2.4 Fraction marginal leak ...................................................................................................... 74
10.2.5 Model for valves, standard flanges and instruments .......................................................... 75
10.2.6 Model for equipment with few incidents .......................................................................... 76
10.3 Validation of PLOFAM .........................................................................................................78
11 Robustness PLOFAM ......................................................................................................................93
11.1 Introduction ........................................................................................................................93
11.2 General evaluation of quality of data basis ..........................................................................93
11.3 Prediction of installation specific leak frequency ..................................................................94
11.4 Quality of the NCS population dataset.................................................................................98
11.4.1 Sensitivity studies with validation model ............................................................................ 98
11.4.2 Adjustment for operational time in equipment counts..................................................... 101
11.5 Time trend and prediction of gas leaks ..............................................................................102
11.6 Randomness ......................................................................................................................105
12 Comparison PLOFAM and SHLFM ................................................................................................107
13 MISOF benchmarking model and PLOFAM...................................................................................118
14 Concluding remarks.....................................................................................................................120
15 Further work ................................................................................................................................122
16 References ...................................................................................................................................124
Appendix A Model Parameters

Appendix B UKCS model: data basis and intermediate results
Appendix C Targets for the PLOFAM model
Technical note no: 105586/TN-6 Rev: Final B Page iii

1 Introduction
This technical note describes:
The methodology for estimation of the initial model parameters based on data from
installations on the UKCS
The validation model based on data gathered from installations on the NCS
The methodology for setting the PLOFAM parameters by use of the validation model
A comparison between PLOFAM and the commonly used model denoted SHLFM.
2 General methodology description

The general methodology is illustrated in Figure 2.1. The method consists of three main steps.
The first step in developing the preferred model parameters has been to establish initial model
parameters solely based on available data gathered from installations on the UKCS. The only data
originating from NCS that has been utilized to perform the parameterization of the initial para-
meters is the relative equipment size distribution taken from the NCS population dataset (see
Chapter 9 in TN-2). This model is denoted the UKCS model. The resulting parameters after
performing the initial parameterization have then been validated in detail towards the data
gathered from installations on the NCS in order to understand whether the model parameters
should be adjusted to improve the models ability to predict the occurrence of leaks on installation
located on the NCS.
In order to analyse the difference in underlying frequency distributions for the different equip-
ment types, a second model alternative is developed where it is attempted to obtain the best
possible fit with the NCS data. This is achieved by running the validation model in an iterative
process. This model is denoted the NCS model.
The ultimate model alternative, denoted PLOFAM, is based on a methodology that weights the
two available datasets. The preferred model is judged to most accurately predict the future
occurrence of leaks with initial leak rate >0.1 kg/s on installations located on the NCS.
Some types of equipment have been subjected to special evaluations where the methodology for
parameterization deviates slightly from the general methodology described above. This is either
due to lack of data in either of the data sets or shortcomings in the available data. The types of
equipment subjected to special evaluations and the reason for the alternative approach is
summarized in Table 2.1.

Step 1: Initial parameterisation
Step 2: Model validation
Establishment of
initial model Step 3: Model development
parameters Validation of
initial model PLOFAM model
based on data Development of
gathered from towards data recommended
installations on gathered from model based on
UKCS (see TN-3) installations on UKCS and NCS
NCS (see TN-2)
data.
Figure 2.1 - Main steps in parameterisation process
Table 2.1 - Equipment types subjected to separate parameterisation process
Equipment Description
type
Hoses No data available in HCRD on leaks occurring under temporary operations

involving use of hoses. The defined model parameters are therefore solely
based on data gathered from installations on the NCS.
Steel pipe The quality of the population data in HCRD for steel pipes is judged to be
poor. Hence, the model is parameterized based on a subset of the NCS
population dataset where equipment counts of length steel pipe are
available. However, available data in HCRD have been used to model the
effect of equipment size on the hole size distribution for steel piping.
Compact No data is available from UKCS and only limited data available from
flanges installations on NCS. A separate assessment is performed to set the model
parameters.

3 Model parameters
The parameters defined in the mathematical formulation of the model are described in TN-5. The
model parameters and additional parameters used in the estimation and validation process are
given in the table below.
Table 3.1 - Summary of all parameters used in the model (see TN-5)
(, ) Hole size frequency distribution (see TN-1) [year-1 component-1]
0 Total leak frequency [year-1 component-1]

F0 = F(d = 1).

FD = F(d = D).
Slope parameter

-1 -1
The fraction of the relevant equipment type with size D

() Probability density function, ()/0
Additional parameters used in validation process
1 Adjustment factor to reflect incidents in the HCR database having a hole size
of 1 mm or less
Historical frequency estimated based on HCR data. See TN-3 and TN-3
Appendix B for frequencies estimated based on HCRD data

4 Model targets
The fundamental assumptions for the model and the requirements to the model are described in
TN-5.
For the ultimate validation of the PLOFAM, the following overall targets have been established:
The model shall reproduce the number of observed leaks in the period 2001-2014 at the
installations in the NCS population dataset (see Chapter 3 in TN-2). The numbers of
installations are 62 and the total number of observed leaks in the period is 181
The model shall reproduce the complementary cumulative leak frequency distribution of the
relevant NCS leaks (181 leaks observed at the 62 installations in the NCS population dataset)
A best possible prediction of the observed number of leaks originating from the various types
of equipment is targeted.
5 Parameterization only based on UKCS data

5.1 Description of methodology
The steps in the parameterization process of the model parameters based on the data from
installations on UKCS (i.e. HCRD described in TN-3) are shown in Figure 5.1.
The process different steps are in practice looped a few times where the criterion for acceptable
fit to the HCR data is based on a manual assessment. A spreadsheet has been developed where
the required data is compiled.
The steps in setting the initial parameters based on HCRD are as follows (per equipment type):
Estimate ()
() describes the trend of total leak frequency with respect to equipment size
Estimate ()
()describes the trend of total leak frequency with respect to equipment size
Estimate 1
This factor is applied to reflect uncertainty with regard to incidents having a hole size around
1 mm.
These steps are run a few times manually (by shifting the model parameters in the spreadsheet)
until an acceptable fit is obtained. There is uncertainty related to this manual assessment, but this
will not affect the ultimate model parameters set based on the validation model that is run
afterwards.
A software routine with a mathematical algorithm for optimization of the parameters was
disregarded based on the following:
It is difficult to define constraints that ensure that the process converge towards a set of
parameters that fit the observed data and at the same time complies with the partly manual
assessments of the
o variation in total failure frequency versus equipment size
o variation in full bore frequency versus equipment size
There is significant uncertainty associated with the UKCS HCR data. The population data has
not been updated after 2006 and there hole sizes are somewhat uncertain (see Chapter 6 in
TN-3)

Ultimately the parameters will be set based on both datasets (UKCS and NCS) applying the
validation model. Hence, to establish full software routine setting the parameters, also the
validation model had to be included in the parameterisation model. The complexity of such a
routine would be considerable without adding any precision to the methodology. The same
ultimate parameters are obtained by applying the validation model and the defined set of
targets (see requirements to model given in TN-5, Chapter 4 and section 10.2.6).
Estimate
1
Estimate
()
Estimate
()
Figure 5.1 - Main steps in parameterisation process based on UKCS data. The process is run
individually per equipment type
5.2 Estimation of total frequency trend with equipment dimension

5.2.1 Methodology
The trend of the total leak frequency is described by the following equation (see Chapter 2 in
TN-5)
0 () = () = (0 0 ) (1)
() is estimated based on:

Qualitative assessment based on properties of the equipment, operational conditions and
typical causes for leaks related to the equipment (which was discussed in a workshop in the
project, Ref. /1/)
Trend described by separate leak frequency calculated for three available equipment size
categories in the HCD database for valves, standard flanges and steel pipes.

() is constrained by the equipment size distribution according to the following:
= 0 (0.25") 0.25" + 0 (0.5") 0.5" + (2)
which in the generalized form for the available data becomes
40" (3)
= 0 ()
=0.25"
The relative equipment size distribution, , is based on the NCS population dataset (see Figure
5.16 and Appendix B). For valves, flanges and steel pipes, is adjusted by weighing the data
extracted from the NCS population dataset with the relative distribution per HCRD equipment
size category (see Figure 5.17).
The only equipment types that have been concluded to possess a trend with equipment
dimension are valves, standard flanges, instruments and steel pipes. For the other types of
equipment, () equals 1.
The resulting () for valves, standard flanges and steel pipe is shown in the following sections.
() for valves is concluded to also apply for instruments.
The uncertainty related to assessment of the parameter values of () is significant. The
uncertainty is controlled by the validation model that will validate that the defined parameters of
() result in an acceptable fit to the observed leaks on NCS installations.
5.2.2 Valve and instruments
Figure 5.2 displays the resulting () for a valve and the corresponding trend with equipment
dimension extracted from HCRD. The equipment size on the first axis for the HCR data is the
weighted mid-point within the HCRD equipment size category based on the equipment size
distribution within this category taken from the NCS population dataset (see Appendix B).
For valves, () is mainly based on the trend observed in the HCR data. The trend for the entire
period (1992-Q1 2015) is considered most reliable. It is has not been identified any casual
arguments in terms of properties of valves or how valves are operated that support the observed
trend. The behaviour of () outside the range of the HCR data (i.e. below about 1.5 and
above about 16) is therefore uncertain. The parameter values are therefore set to avoid large
shift with increasing equipment dimension, but at the same time reproducing the observed trend
in the HCR data. Due to the lack of casual arguments, the uncertainty related to () for valves
is considered to be more prominent than for standard flanges and steel pipes (see following
sections).
(D) for valves is judged to apply also for instruments. Small valves are often associated with
instruments, which is probably a major contributor to leaks from instruments. Furthermore, the
range of equipment sizes for instruments is small, and the effect of an imprecise model is small.

Figure 5.2 Valves (total of manual and actuated valves): () as function of equipment diameter,
D. The parameters used to determine () are given in Appendix A
5.2.3 Standard flange

Figure 5.3 displays the resulting () for a standard flange and the corresponding trend with
equipment dimension extracted from HCRD. The equipment size on the first axis for the HCR
data is the weighted mid-point within the HCRD equipment size category based on the
equipment size distribution within this category taken from the NCS population dataset (see
Appendix B).
() is to a large degree based on the observed trend in the HCR data. The trend for the entire
period (1992-Q1 2015) is considered most reliable. This trend can partly be explained by the
causes for leaks originating from flanges. Most flange leaks are related to latent errors caused by
human errors during work on the equipment (e.g. maintenance, installation of equipment). The
distribution of leak at installations on the NCS with respect to cause is shown in Figure 5.4
(Ref. /2/). The typical latent errors related to interventions are:
Too low/high moment when tensioning
Wrong (or no) gasket used
In general not following procedure when installing the flange.

It is judged that such latent errors are more likely for large equipment than small equipment
because:
Larger flanges are more difficult to handle
Larger flanges require more moment when tensioning the bolts
A larger flange surface is more challenging mount correctly.
This factors support a moderate positive trend in () for increasing equipment dimension,
which is consistent with the observed trend in the HCR data.
The behaviour of () outside the range of the HCR data (i.e. below about 1.5 and above ab-
out 16) is considered more uncertain than within the range of the HCR data. The parameter
values are therefore set to avoid large shift with increasing equipment dimension, but at the
same time reproducing the observed trend in the HCR data.
Figure 5.3 - Standard flanges: () as function of equipment diameter, D. The parameters used to
determine () are given in Appendix A

Figure 5.4 - Standard flanges: Leaks at installations on the NCS classified according to cause
(Ref. /2/)
5.2.4 Steel pipe

Figure 5.5 displays the resulting () for a steel pipe and the corresponding trend with
equipment dimension extracted from HCRD. The equipment size on the first axis for the HCR
data is the weighted mid-point within the HCRD equipment size category based on the
equipment size distribution within this category taken from the NCS population dataset (see
Appendix B).
() is to a large degree based on the observed trend in the HCR data. The trend for the entire
period (1992-Q1 2015) is considered most reliable.
Based on the observed causes for leaks from steel pipes, it is considered reasonable that the HCR
data displays a decreasing trend with increasing equipment dimension. Causes for leaks
stemming from pipes are dominated by technical errors such as corrosion, failure of welds and
fatigue (Ref. /2/). The wall thickness of pipes increases in general with increasing equipment
dimension, which implies that the big equipment is more robust than small pipes. This argument
also applies for external impacts.
The behaviour of () outside the range of the HCR data (i.e. below about 1.5 and above
about 16) is considered more uncertain than within the range of the HCR data. The parameter
values are therefore set to avoid large shift with increasing equipment dimension, but at the
same time reproducing the observed trend in the HCR data.

Figure 5.5 - Steel pipe: () as function of equipment diameter, D. The parameters used to
determine () are given in Appendix A
5.3 Full bore hole fraction

5.3.1 Methodology
The trend of the total leak frequency is described by the following equation (see Chapter 2 in
TN-5)
() = 0 () () = 0 () ( + ) (4)
The rupture fraction, (), is estimated based on:

Qualitative assessment based on properties of the equipment, operational conditions and
typical causes for incidents related to the equipment (which was discussed in a workshop in
the project, Ref. /1/)
The full bore hole fraction per equipment dimension for valves, flanges and steel pipes (the
data is enclosed in Appendix B) based on the reported equipment dimension and hole size
per leak in HCRD

Approximation of full bore hole fraction estimated from HCR data (see Figure 5.6) for other
equipment types than valves, flanges and steel pipes.
The uncertainty related to assessment of the parameter values of () is prominent. However,
the concluded shape of () is supported by the expected behaviour of () for the various
equipment types. Secondly, the validation model will verify that the defined parameters of ()
result in an acceptable fit to the observed leaks on NCS installations.
() is to a large degree based on the observed full bore hole fractions extracted from HCRD.
The HCR data display a considerable difference for the period 1992-Q12015 and 2001-Q2015,
which implies that the fraction medium and large leaks (> 1 kg/s) is less in the period after year
2000. This is consistent with the relative complementary leak rate distributions extracted from
HCRD shown in Figure 5.7 through Figure 5.9 (see TN-3 for further description). The relative
complementary leak rate distributions for leaks at installations on the NCS is however more
similar to the distribution for the period the period 1992-Q12015 than the period after 2000.
The reason for this result has not been fully understood. It may be due to inaccuracies in the
parameters logged in HCRD (see Chapter 6 in TN-3 as well as Chapter 0, where the difference in
relative complementary leak rate distributions for installations on the NCS and the UKCS is
discussed further).
Figure 5.6 Average rupture fraction for all equipment sizes for various equipment types based on
an assessment of leak judged to represent full bore hole in HCRD. The data is based on Filter 1 (a
and b respectively). A hole is interpreted as being a full bore hole if the hole size is within 2%
deviation from a whole number when dividing the hole size with 25.4 mm/inch

Figure 5.7 Relative complementary cumulative frequency leak rate distributions for significant
leaks for installations on the UKCS for the period 1992-Q1 2015 and 2001-Q1 2015. Note that
incidents are extracted with respect to estimated initial leak rate and not the hole size
Figure 5.8 Relative complementary cumulative frequency leak rate distributions for installations
on the UKCS for the period 2001-Q1 2015. Note that incidents are extracted with respect to
estimated initial leak rate and not the hole size

on the UKCS for the period 1992-Q1 2015. Note that incidents are extracted with respect to
5.3.2 Valves
() for valves (manual and actuated valves altogether) is displayed in Figure 5.10 together with
available data from HCRD. The underlying data of full bore hole fraction is reported in detail in
Appendix B.
The trend for the entire period (2001-Q1 2015) is judged to be most representative for the period
after 2000 for installations on the NCS. Both datasets, i.e. both time periods, do however
demonstrate a decreasing trend with increasing equipment dimension.
The decreasing trend can partly be explained by effects related to the causes for leaks originating
from valves. A major fraction of valve leaks are related to latent errors caused by human errors
during work on the equipment (e.g. maintenance, installation of equipment). The distribution of
leaks at installations on the NCS with respect to cause is shown in (Ref. /2/). The typical latent
errors related to interventions are:
Valve left in wrong position after the operation (e.g. maintenance)
Valve erroneously installed.
Another relevant failure model in this regard is valves erroneously operated during normal
operation.
In general, it is judged that small valves are being operated more frequently than large valves.
Combined with that procedures are more stringent for operation of large valves compared to
small valves, it is judged that the full bore hole fraction should decrease with increasing
dimension of the valve.

Although identified casual arguments and data from HCRD are consistent, the uncertainty related
to assessment of the specific parameter values of () for valves is significant. However, the
uncertainty is reduced considerably by the validation model. The validation model will verify that
the defined parameters of () result in an acceptable fit to the observed leaks originating from
valves at installations on the NCS.
Figure 5.10 - Valve: () as function of equipment diameter, D. The full bore hole fraction
extracted from HCRD is enclosed in Appendix B

Manual valve leaks
20
18
16
14
12
10
8
6
4
2
0
Technical Latent errors Immediate Process Design External Unknown
(N.O.) (O) (O) (N.O.) (N.O.) (N.O.) (N.O.)
Figure 5.11 - Manual valves: Leaks at installations on the NCS classified according to cause (Ref. /2/)
Actuated valve leaks

18
16
14
12
10
0
Technical Latent errors Immediate Process Design External Unknown
(N.O.) (O) (O) (N.O.) (N.O.) (N.O.) (N.O.)
Figure 5.12 Actuated valves: Leaks at installations on the NCS classified according to cause
(Ref. /2/)

5.3.3 Standard flange
() for standard flanges is displayed in Figure 5.13 together with available data from HCRD.
The underlying data of full bore hole fraction is reported in detail in Appendix B.
after 2000 for installations on the NCS. Both datasets, i.e. both time periods, do however
demonstrate a decreasing trend with increasing equipment dimension. The decreasing trend in
full bore hole size as a function of equipment size is somewhat less than for valves.
from standard flanges. A major fraction of leaks from flanges are related to latent errors caused
by human errors during work on the equipment (see Figure 5.4).
The procedures are more stringent for operation of large flanges compared to small flanges. It is
also considered less likely that a large flange is mounted erroneously compared to small flanges.
The number of bolts increases with increasing flange size. The criticality of improper tightening of
one bolt (or a couple) becomes less critical with increasing number of bolts. it is also judged that
erroneous mounting is more easily visually observed by the operator for a large flange compared
to a small flange.
to assessment of the specific parameter values of () for standard flanges is significant.
However, the uncertainty is reduced considerably by the validation model. The validation model
will verify that the defined parameters of () result in an acceptable fit to the observed leaks
originating from standard flanges at installations on the NCS.
Figure 5.13 - Standard flange: () as function of equipment diameter, D. The full bore hole
fraction extracted from HCRD is enclosed in Appendix B

5.3.4 Steel pipe
() for steel pipes is displayed in Figure 5.14 together with available data from HCRD. The
underlying data of full bore hole fraction is reported in detail in Appendix B.
after 2000 for installations on the NCS. Both datasets, i.e. both time periods, do however dem-
onstrate a decreasing trend with increasing equipment dimension. The decreasing trend in full
bore hole size as a function of equipment size is equivalent with the trend for valves.
from steel pipes, which is described in section 5.2.4. The casual arguments for a decreasing total
frequency for leaks with equipment dimension for steel piping is judged to apply also for the full
bore hole size fraction. The general casual argument is that the robustness of the pipe increase
with increasing equipment dimension.
to assessment of the specific parameter values of () for steel pipe is significant. However, the
uncertainty is reduced considerably by the validation model. The validation model will verify that
the defined parameters of () result in an acceptable fit to the observed leaks originating from
steel pipes at installations on the NCS.
Figure 5.14 Steel pipe: () as function of equipment diameter, D. The full bore hole fraction
extracted from HCRD is enclosed in Appendix B

5.4 Uncertainty small hole sizes
1 is applied to reflect uncertainty with regard to incidents having a hole size around 1 mm.
The exact hole size for these incidents logged in HCRD is judged to be particularly uncertain.
1 is set based on an assessment of the fit towards the complementary cumulative hole size
frequency distribution extracted from the HCRD (see Figure 5.18 for valves and Appendix B for all
equipment types). The actual values established for 1 are given in Table 5.1. The assessment
is based on a manual evaluation.
The upper threshold is defined by the parameter value obtained from the HCR data, which is
presented in Figure 5.15.
1 relates to as follows:
= 1 (5)
As the uncertainty related to the hole size close to 1 mm is high, the 1 factor is also
uncertain. 1 affect the slope of the distributions, i.e. the parameter . The importance of
1 for the ultimate PLOFAM parameters is however negligible. The PLOFAM parameters is set
based on a methodology where a best fit towards the NCS data is targeted, which overrule the
relatively small effect 1 has on the UKCS model alternative. For most equipment types, the
parameter value of 1 is set exactly equal to the observed value in Figure 5.15. The values
used are given in Table 5.1.
A hole with a diameter less than 1 mm does only in extreme conditions generate an initial leak
rate above 0.1 kg/s. For the process conditions at installations in the NCS population dataset (see
TN-5), holes less than 1.5 mm does not contribute to leaks having an initial leak rate above 0.1
kg/s. Hence, the shape of the frequency distribution in the region around 1 mm is not important
in QRAs where the leak rate generating significant consequences is 0.1 kg/s or higher. In small
poorly ventilated modules, smaller leak rates may be relevant to consider, and the contribution
from holes with a diameter less than 1 mm should be considered. Neither the data in HCR nor
the data gathered from installations at the NCS can be used to conclude on the behaviour of the
distribution for such small holes. It may be that the slope of the distribution changes when the
hole size approaches zero opposed to the slope of the distribution for bigger holes. However, the
parameterized mathematical formulation applied has demonstrated capability to reproduce the
observed distribution of leaks having an initial leak rate equal to or larger than 0.1 kg/s to the
release rate generated in case of rupture of the equipment (i.e. full bore hole). This is an
argument for that the distribution also are able to represent the underlying behaviour of the
distribution of holes having a diameter less than 1 mm. Further investigations, and preferably
access to data of leaks stemming from holes having a diameter less than 1 mm, is required to
investigate further.

Figure 5.15 - 1 based on HCRD for various equipment types for the period, i.e. the total leak
frequency divided by leak frequency for hole size >1 mm. The figure is taken from TN-3, where
corresponding numbers for Marginal leaks also are given
5.5 Fitting to observed frequency hole size distribution from HCRD

The last step is an manual assessment of the fit to the complementary cumulative hole size
frequency distribution extracted from the HCRD for the various equipment types.
The objective is to estimate () per equipment type per equipment size given by the following
equation (see TN-5)
() = 0 (6)
The data extracted by use of Filter 1b, see TN-3, is in general used to assess the fit of the model
parameters. In Filter 1, all relevant process leak incidents in the HCR-data are included except:
Incidents recorded with hole size N/A.
The applied filters and period for calculation of frequency is summarized in Table 5.1. The only
equipment type where the period 1992-Q1 2015 is used to set the frequency is flexible piping,
which is due to the scarce data available for the period 2001-Q1 2015.
The total () for all equipment sizes for an equipment type, denoted (), is obtained by
weighted the individual () per equipment size with the fraction of equipment with equipment
dimension , denoted . This can be expressed as follows:
() = 0.25" () 0.25" + 0.5" () 0.5" + (7)

which in the generalized form for the available data becomes
40" (8)
() = ()
=0.25"
The results for valves and standard flanges are shown in Figure 5.18 and Figure 5.19 respectively.
The curve named Total, all equipment sizes (dotted green curve) represents FTot (d). The
distribution with black bold font (UKCS Historical, Valve) is the observed () extracted
from HCRD, which is dependent on the actual distribution of equipment size on installations on
the UKCS.
The equipment size distribution for equipment on installations on UKCS is probably somewhat
different from installations on the NCS. This adds uncertainty to the parameterisation
methodology as mainly is based on data from installations on the NCS. The relative difference
for the equipment dimension categories available in HCRD for valves, see Figure 5.17, indicates
that equipment are on average slightly bigger at installations located on the NCS compared to
installations on the UKCS. This is may be explained that NCS installations process (on average)
more gas than UKCS installations. Equipment containing gas tends to be bigger than equipment
processing liquid (see Chapter 9.6 in TN-2). In any case, the uncertainty related to equipment size
distribution on the ultimate model is considered negligible. PLOFAM is in any case validated
thoroughly towards a comprehensive population dataset that applies to installations on the NCS.
The weighted contribution from the individual equipment dimensions (i.e. the term () in
equation (8)), are illustrated with thin font with an individual colour for each equipment
dimension in Figure 5.18.
The results for all equipment types are shown in Appendix B. The resulting probability density
function (denoted () = ()/0 per equipment dimension per equipment type) for valves is
shown in Figure 5.23. The resulting () per equipment dimension for valves is shown in Figure
5.22. () and () for other equipment types can be studied by applying the parameters
enclosed in Appendix A.
A major uncertainty for some types of equipment is the low number of leaks in the dataset. This
applies in particular to filters, process vessels (including atmospheric vessels), pumps, com-
pressors, pig traps and heat exchangers (all types). Figure 5.20, which is the result for centrifugal
compressors, demonstrates that extrapolation outside the domain where data is available is
required to parameterize the model for some types of equipment. The challenge is the same also
if the whole observation period is used and all types of compressors are used as basis (see Figure
5.21). Hence, the obtained parameters for the mentioned types of equipment must be seen in
light of the scarcity of the available data. Therefore, a separate methodology has been developed
to set the parameters for these equipment types where the data from UKCS and NCS are pooled
(see Chapter 10.2.6).
The resulting model parameters are enclosed in Appendix A. The model based solely on HCRD is
denoted the UKCS model.

Figure 5.16 - for valves: Distribution on various equipment dimensions (see Appendix B for
detailed data)

Figure 5.17 - Valves: Relative distribution HCRD equipment size categories for exposure time
extracted from HCRD (see TN-3) and NCS population dataset (see TN-2)
Table 5.1 Filters and time periods used for each equipment type to estimate initial parameters
solely based on HCRD (see TN-3)
Equipment type Filter used for Time period for K1mm
hole size estimation of F_hist
distribution
Air-cooled heat exchanger 1) 1992-Q12015 NA
NA
Atmospheric vessel 1a 2001-Q12015 1.00
Centrifugal compressor 1b 2001-Q12015 1.25
Centrifugal pump 1b 2001-Q12015 2.40
Compressor 1b 2001-Q12015 1.22
Filter 1b 2001-Q12015 1.00
Flexible pipe 1a 1992-Q12015 1.28
Gas lift well 1a 2001-Q12015 3.00
Instrument 1b 2001-Q12015 1.16
Pig trap 1a 2001-Q12015 1.21
Plate heat exchanger 1b 2001-Q12015 1.00
Process vessel 1a 2001-Q12015 1.30
Producing well 1a 2001-Q12015 1.30
Pump 1b 2001-Q12015 1.33
Reciprocating compressor 1b 2001-Q12015 1.20
Reciprocating pump 1b 2001-Q12015 1.33

Equipment type Filter used for Time period for K1mm
hole size estimation of F_hist
distribution
Shell and tube heat 1b 2001-Q12015 1.10
exchanger
Shell side heat exchanger 1b 2001-Q12015 1.00
Standard flange 1b 2001-Q12015 1.00
Steel pipe 1b 2001-Q12015 1.31
Tube side heat exchanger 1b 2001-Q12015 1.00
Valve 1b 2001-Q12015 1.51
Manual valve 1b 2001-Q12015 1.68
Actuated valve 1b 2001-Q12015 1.38
1) Only one relevant leak. Hole size distribution parameters assumed to be equivalent with Shell and tube
heat exchanger
Figure 5.18 - Valve: resulting fit to the complementary cumulative frequency hole size distribution
extracted from the HCR database (based on filter 1b, see TN-3). The curve denoted Total, all
equipment sizes represents the weighted () for all equipment sizes, denoted () (see
equation 8)

Figure 5.19 Standard flange: resulting fit to the complementary cumulative frequency hole size
distribution extracted from the HCR database (based on filter 1b, see TN-3). The curve denoted
Total, all equipment sizes represents the weighted () for all equipment sizes, denoted ()
(see equation 8)

Figure 5.20 - Centrifugal compressor: resulting fit to the complementary cumulative frequency hole
size distribution extracted from the HCR database (based on filter 1b, see TN-3). The curve denoted
(see equation 8)

Figure 5.21 Compressors (total of centrifugal and reciprocating compressors): resulting fit to the
complementary cumulative frequency hole size distribution extracted from the HCR database
(based on filter 1a, see TN-3). The curve denoted Total, all equipment sizes represents the
weighted () for all equipment sizes, denoted () (see equation 8)

Figure 5.22 - Valves: resulting () for valves for parameters developed solely based on HCRD. This
model is denoted the UKCS model

Figure 5.23 - Valves: resulting probability density function (denoted ()) for valves for parameters
developed solely based on HCRD. This model is denoted the UKCS model
5.6 Shape of the hole size frequency distribution

Under development of the mathematical framework described in TN-5, an alternative model
including a constant term, denoted 1 , was outlined. The alternative model is expressed as follow
() = (0 1 ) + 1 (9)
The terms are linked through the parameter alpha according to the following equation
1 (10)
=

can have a value in the interval 0 1.

In the parameterisation process, this alternative model where also considered. Although the
alternative model is believed better describe the underlying general probability density function
for holes, it has been concluded that the additional term does not increase the ability of the
model to describe the observed frequency for leaks at installations on the UKCS and NCS. This
indicates that the properties of the underlying mathematical distribution have not been fully
understood. This should be further investigated in future projects when more data is available.
The only equipment type where the HCR data strongly indicates that a constant term should be
included is standard flange. This is based on that () (see equation 8) for standard flanges
extracted from HCRD displays a decreasing slope with increasing hole size. The resulting model
for standard flanges using an equal to 0.9 is shown in Figure 5.24. The results show that the fit
to the HCR data is somewhat better than without the constant term, which is shown in Figure
5.19.

The validation model (LeakPy) results for all model alternatives, including PLOFAM, demonstrate
that the model for standard flanges without the constant term (i.e. set to zero) is able to the
reproduce the complementary cumulative distribution of leaks at NCS originating from standard
flanges. This is shown in Figure 9.14 and Figure 10.15 for the NCS model and PLOFAM
respectively. Hence, it is concluded to disregard the alternative model with the constant term.
The resulting model will become less complex without affecting the performance of the model
significantly.
Figure 5.24 Standard flange: resulting fit to the complementary cumulative frequency hole size
distribution extracted from the HCR database (based on filter 1b, see TN-3). The curve denoted
(see equation 8). Alternative model of () described by equation 9 and 10 is used in this case.
5.7 Modelling of various valve types

The historical leak frequencies (frequency for holes 1 mm) for various types of valves and the
equipment size categories available in HCRD are shown in Figure 5.25 through Figure 5.30. The
results show that the trends are similar for the two time periods, i.e. 1992-Q1 2015 and
2001-Q1 2015.
The results show that there is a considerable difference between actuated valves and manual
valves. It has not been identified any casual arguments for the significant difference between
manual and actuated valves.
In the general, the physical properties of manual and actuated valves are similar. The main
difference is the how the valve is operated. In fact, it is considered reasonable to argue that the
frequency for leak from actuated valves should be less than the frequency originating from

manual valves because actuated valves are less frequently subjected to manual interventions (in
particular for small valves). The term manual includes in this context also remote operation.
Weight has been put on the HCR data when setting when setting the parameters for () and
() without support by casual arguments. The situation is however unique for valves. The
difference in frequency for leaks stemming from manual and actuated valves is counterintuitive
and no casual arguments have been identified. it is therefore concluded to recommend to use
the amalgamated frequency for manual and actuated valves for all types of valves in PLOFAM.
Specific model parameters for actuated and manual valves have still been developed, and can be
found in Appendix A. The parameters can be used in special cases, but must be used carefully
acknowledging the uncertainty described in this chapter. An example of a case where the use of
the specific model for manual and actuated valve may be beneficial is studies that address the
risk reducing effect of ESD valves in terms of reduction in inventory (which result in reduced fire
duration and a reduced exposure time of combustible gas mixed with air to potential ignition
sources). The results show that the frequency for leaks from ESD valves is significantly less than
the frequency for leaks from originating from other actuated valves. In such cases, the type of all
valves must be known and counted. The reduction factor for actuated ESD valves relative to other
actuated valves is set to 0.6. This gives a reduction in of 40 % according to the following
, = , 0.6 (11)
Figure 5.25 HCRD period 1992-Q1 2015: leak frequency (for holes 1 mm) per valve year for
different types of actuated valves per equipment category ( 3, 3-11 and > 11)

different types of actuated valves per equipment category ( 3, 3-11 and > 11)
different types of valves

different types of valves
actuated and manual valves per equipment category ( 3, 3-11 and > 11)

actuated and manual valves per equipment category ( 3, 3-11 and > 11)
6 Description validation model

6.1 Methodology
The various model alternatives are systematically validated by use of the software routine
denoted LeakPy. LeakPy calculates the following for each of the installations in the NCS
population dataset and the full NCS population dataset (see Chapter 9 in TN-2):
The frequency distribution versus initial leak rate per equipment type
The estimated number of leaks in the relevant period (base case is 01.01.2001-31.12.2014)
versus initial leak rate per equipment type.
The results are compiled and presented along with the corresponding observed leaks in the two
population datasets (NCS population dataset and full NCS population dataset).
A subset of the plots used to assess the model performance is presented in the chapters
discussing the various model alternatives.
The contribution from significant and marginal leaks are calculated an individually. Results are
thus available for each of leak scenario separately. The validation model does however focus on
validating the total of marginal and significant leaks. This is because the total of significant and
marginal leaks is considered to be more reliable in the NCS population dataset than the individual
contribution per scenario type. The fraction marginal leaks found in the NCS dataset are judged
to be somewhat low (see Chapter 5 in TN-2).
The contribution from steel piping is in the validation model estimated by adding a relative contri-
bution based on the total frequency estimate for the other equipment types. This is because
equipment counts of steel pipe length are only available for 24 installations in the NCS
population dataset (see Chapter 9.3 in TN-2). The equipment counts for those 24 installations are
utilized to estimate for steel pipe based on the NCS data (see Chapter 8.3).

The contribution from flexible piping and hoses are grouped together in the plots generated from
the validation model. However, the contribution from flexible piping is marginal as the number of
meters of flexible piping in the population dataset is small.
A separate category for storage tanks is shown in the plots. However, no equipment counts or
model for the leak frequency for such equipment has been developed.
6.2 Release modelling

6.2.1 General
Equations for calculating initial gas and liquid releases are given in this chapter for full pressure
leaks and zero pressure leaks. The initial release rate for a marginal leak is modelled in the same
way as a full pressure leak.
6.2.2 Gas releases
Choked conditions occur as the downstream pressure falls below a critical value . That critical
value can be calculated from the dimensionless critical pressure ratio equation:
2 (12)
= 1
0 + 1
Assuming =1 bara and = 1.31 for Methane (see Table 6.2), gives 0 =1.8 bara. Thus gas
releases from inventories with over pressure >0.8 barg, which in most cases is the situation for
process leaks, should be modelled using the equation for chocked mass flow rate given by the
following relationship:
+1 (13)
2 1
= 0
0 ( + 1)
where the parameters in the equation are given in the table below. Rearranging the above and
g M
noting that = gives:
P0 RT0
+1 (14)
2 2 1
= 0
4 ( + 1) 0
The molar mass is not given in the EQCDB. For validation purposes it is assumed that the leaking
gas is Methane. Relevant specific heat ratios are given in Table 6.2. Substituting = 1.31 for
methane, = 0.85, converting the units of pressure to bara and noting that the units of the
diameter are in mm we have:
1.31+1 (15)
2 1.311 105
= 0.85 1.31 2 ( + )
4 (1.31 + 1) 106
Giving:
(16)
= 1.41246 104 2 ( + )
Where is the hole size [mm], is the initial gas density [kg/m] and is the initial gas pressure
[barg] and is the atmospheric pressure.

Note that full pressure gas releases from inventories with over pressure < 0.8 barg should have
been modelled using equations for compressible flow. However, in the validation process only
the equation for chocked flows is used.
Table 6.1 - Parameters used to calculate gas release rates
Initial gas release rate [kg/s ]
Discharge coefficient =0.85
Hole area [mm]
Hole diameter [m]
Hole diameter [mm]
0 Initial gas pressure [Pa]
Initial gas density [kg/m]
M Molecular weight of gas [kg/mol]
Specific heat ratio (see Table 6.2)
Universal gas constant = 8.314 J/(K mol)
0 Initial gas temperature [K]
Initial gas pressure [barg]
Atmospheric pressure [bara]
Table 6.2 - Specific heat ratio for relevant gases
Gas Specific heat ratio,
Methane 1.307
Propane 1.131
Buthane 1.096

6.2.3 Liquid releases
Pure liquid releases are modelled as incompressible fluid flows. Thus the following relationship
applies:
(17)
= 2 (0 ) + 22
where the parameters in the equation are given in Table 6.3. By neglecting the liquid head, h (see
the effect this has in Figure 6.1), substituting = 0.61, converting the units of pressure to bar,
noting that the units of the diameter are in mm and replacing the pressure term with the gauge
pressure of the liquid, this can be simplified to:
(18)
105 2
= 0.61 2
4 106
giving
(19)
= 2.14257 104 2
Figure 6.1 - The effect of neglecting liquid head. The figure gives the fraction of the actual leak
rate calculated by neglecting the liquid head h, for a range of values for the liquid head
Table 6.3 - Parameters used to calculate gas release rates
Initial liquid release rate [kg/s ]
Discharge coefficient =0.61
Hole area [mm]
Hole diameter [mm]

Liquid head [m]
Gravitational acceleration [m/s]
0 Initial liquid gas cap pressure (inventor pressure) [Pa]
Atmospheric pressure [Pa]
Initial liquid density [kg/m]
Initial liquid gas cap pressure (inventory pressure) [barg]
7 Validation of model based on UKCS data

The model validation results for the model parameters solely based on HCR data are shown in
Figure 7.1 through Figure 7.18. This model alternative is denoted the UKCS model.
The main properties of this model alternative are shown in Table 7.1.
Table 7.1 - Main properties of the UKCS model properties
Item Description
Basis for estimation of model Solely based on HCRD data (see Chapter 5.5)
parameters
Equipment types not included Model for hoses not included
Model for steel piping The contribution from steel pipes is modelled by adding a
contribution that equals the historical relative fraction
originating from steel piping. The historical average for gas
and liquid leaks in the period 2001-Q12015 from HCRD,
which is 24%, is applied. This fraction applies for the total
frequency after the contribution from steel pipe has been
added. This means that the total frequency without steel
pipe are to be increased with the factor
1
1.316
(10.24)
The results show that the fit to the NCS data with regard to the number of leaks (see Figure 7.3
and Figure 7.6) is quite good. The estimated number of leaks is 216, whilst the number of ob-
served leaks is 181, which constitute a deviation of about 20 %. The contribution from tempo-
rary operations involving use of hoses is not included. The contribution from flexible piping and
hoses are grouped together in the plots generated from the validation model. Hence, the small
contribution from hoses seen in Figure 7.9 and Figure 7.10 is related to flexible piping
permanently in operation.. The number of leaks stemming from hoses is 11 in the NCS
population dataset. This means that the actual deviation between the total number of leaks
predicted by the UKCS model and the 181 leaks observed at NCS is somewhat larger than 20 %
(25 % rather than 20 %).

Figure 7.4 and Figure 7.5 shows that the model will on average slightly overpredict the number
of observed leaks on the installations in the NCS population dataset. A perfect model would
produce a prediction of leaks that is located on a line where the number observed leaks equals
the predicted number of leaks (the red dashed line in Figure 7.4 and Figure 7.5). The general
explanation for the observed variance is deficiencies related to the modelling principle and
randomness, which is further explained in Chapter 11.
Overall, the model tends to underpredict the observed number of leaks for installations that has
experienced a high number of leaks, and overpredict the number of leaks for installations where
few leaks have been observed in the period (01.01.2001 31.12.2014). This is to a large degree
explained by the decreasing trend in number of observed leaks with time throughout the period
2001-2014. The number of leaks has decreased with time for all installations in the dataset. The
decrease is however most prominent for the installations being in production the entire period.
These installations tend also to be somewhat larger than the newer installations. This is the in-
stallations where the greatest number of leaks is observed, and is the installations the model
tends to underpredict. The decreasing trend with time with respect to observed leaks is most pro-
minent in the period 2001-2006 (see Chapter 10 in TN-2). The model performance per
installation is therefore better for the time period after 2006. This is demonstrated in in Chapter
11.
The complementary frequency distributions with respect to initial leak rate, shown in Figure 7.9
and Figure 7.10, displays that distribution with respect to initial leak rate is skewed towards
larger leak rates relative to the observed data from the NCS. The results show that difference
between the predicted and observed complementary frequency distribution emerges for leaks
around 1 kg/s, and increase with increasing initial leak rate.
Figure 7.1 and Figure 7.2 shows the complementary frequency distributions versus hole size for
all equipment types. The hole size is considered to be a better parameter than the initial leak rate
when comparing data from UKCS (i.e. HCRD) and NCS. The observed inconsistency between the
hole size and calculated initial leak rate for the leaks recorded in HCRD (see chapter 6 in TN-3)
has not been explained, but it is judged that hole size is the most reliable parameter. The
frequency distribution, Figure 7.1, shows the same trend as the complementary distributions with
respect to initial leak rate (i.e. Figure 7.9. The difference between the predicted frequency
distribution based on the UKCS model parameters and observed UKCS data emerges around 10
mm, and increases for larger hole sizes. The normalized complementary frequency distributions
(i.e. Figure 7.10 and Figure 7.2) also display the same trends, but the difference is more apparent
for the distribution with respect to hole size. This is partly caused by that the model has
accounted for the uncertainty related to incidents with a hole size around 1 mm through the
introduction of the parameter denoted 1 in the validation process. The difference in the total
accumulated frequency between the distribution for installations on the NCS and the observed
distribution extracted from HCRD equals the average value of 1 displayed in Figure 5.15 (i.e.
about 1.25). The remaining deviation may be explained by underreporting of leaks having a hole
size less than 1 mm and/or that the parameterized distributions does not represent shape of the
underlying distribution for small hole sizes.
This deviation between the distributions may be partly explained by the difference in type of
equipment (e.g. number of flanges per installation at UKCS vs. NCS) and dimension of
equipment at installations on the NCS versus installations on the UKCS. The comparison of
equipment dimension for valves shown in Figure 5.17, demonstrated that there is a difference in
equipment size that will affect the result. The difference in the complementary frequency
distribution with respect to initial leak rate will also be affected by the difference in equipment
properties in addition to difference in process conditions (i.e. composition, pressure and
temperature of fluids being processed). It is hard to assess the total effect of difference in
equipment and process conditions between the installations at NCS and UKCS. The most
important factor is however considered to be uncertainty related to the HCR data (se chapter 6 in
TN-3). Overall, the uncertainty related to the HCR data and difference in physical properties of
installations at UKCS and NCS may explain the observed difference between UKCS and NCS.

In addition to the overall uncertainty related to population data and the provided hole sizes in
HCRD, which is discussed in chapter 6, the uncertainty related to the data basis for assessment of
() (see Figure 5.6) obtained from HCRD, is also important. Figure 5.6 shows that the rupture
fraction is prominent for many equipment types, which has resulted in a quite high rupture
fraction for some equipment types. This applies in particular to compressors, heat exchangers and
process vessels. Figure 7.10 shows that the contribution to large leaks (> 10 kg/s) is significant for
these equipment types. The uncertainty related to parameterisation of () may partly explain
the difference in the frequency for big holes / large initial leak rates between the UKCS model
prediction and observed data from installations at the NCS.
The simplified modelling of the contribution from steel pipe amplifies the effect on large leaks
due to the average fraction is so prominent (24 %). Furthermore, the relative contribution from
steel piping taken from the HCRD is judged to somewhat uncertain (see TN-3). In other words, it
is judged that a more accurate estimate of the leak frequency per meter steel pipe combined
with exact population for all 62 installations in the NCS population dataset would result in a
smaller predicted frequency for large leaks (> 1 kg/s) stemming from steel piping.
Figure 7.7 and Figure 7.8 displays that the UKCS model is not able to predict the relative contri-
bution for the different types of equipment very well. There is a prominent deviation for most
equipment types. It is judged that this to large extent can be explained by the factors mentioned
above, i.e. uncertainty related to the HCR data and difference in physical properties of installa-
tions at NCS and UKCS (i.e. equipment types, equipment dimensions and process conditions).
One more factor not previously mentioned is the difference in categorisation of leaks stemming
from valves and flanges related to instrument connections. A leak at NCS stemming from a
valve/flange associated with an instrument connection is classified as a leak originating from a
valve/flange. In HCRD, most such leaks are probably attributed to the instrument connection. The
way the population data in HCRD and the NCS population data are established for instruments
are however expected to be quite consistent. This means that the observed deviation between
relative contribution from valves, standard flanges and instruments can be explained by diffe-
rence in how leaks from these equipment types have been classified in terms of equipment type.
The complementary frequency distributions with respect to initial leak rate, shown Figure 7.17
and Figure 7.18, displays that relative contribution from significant leaks and marginal leaks is not
representative for the installations at NCS. As described in TN-2 (see Chapter 5 in TN-2), it is
judged that the fraction of marginal leaks in the NCS dataset is somewhat smaller than the actual
underlying fraction. Figure 7.17 shows that the difference between the observed and predicted
distribution for significant leaks is considerable less than the same deviation for marginal leaks.
The big deviation for marginal leaks affects the results for the total leaks seen in Figure 7.1
through Figure 7.16. The observed difference between UKCS and NCS installations with respect
to leak scenario is judged to mainly be due to shortcomings of the NCS data, i.e. the actual
number of marginal leaks is expected to be higher than documented by the NCS dataset.
Accounting for the factors that may explain the observed difference between the UKCS model
and the behaviour of observed leaks at NCS installations, the deviation between the observed
data at NCS and model prediction generated by the UKCS model is considered to be small. The
overall good fit is interpreted as an strong indication for that the underlying frequency for leaks
at installation located on the NCS and the UKCS are similar. This conclusion is supported by the
complementary cumulative distribution of leaks per equipment type shown in Figure 7.13
through Figure 7.16. The trend is similar for all of the equipment types that constitute the major
fraction of leaks. The main difference is the number of aggregated leaks for valves and steel
piping, which also can be seen in Figure 7.7 and Figure 7.8. For other equipment types (e.g.
pumps, compressors, vessels and heat exchangers) the complementary cumulative distribution of
leaks per equipment type cannot be illustrated because the number of incidents in the NCS
population dataset is too low for these equipment types.
An apparent weakness of the UKCS model illustrated by the validation model is the inadequate
ability to predict the relative contribution from liquid and gas leaks. Figure 7.11 and Figure 7.12
shows that the model underpredict the number of small gas leaks (< 1 kg/s), but overpredict the

number of all liquid leaks considerably. The model deficiency in prediction of the relative contri-
bution from gas and liquid leaks is also an attribute of the SHLFM model commonly used in
industry in Norway (see Figure 12.11 and Figure 12.12). This is shortcoming of the model is
discussed further in Chapter 10.3.
Overall, the validation model shows that the UKCS model does not entirely reproduce the
observed leaks at installations on the NCS. The difference is small, but it is considered that an
improved model alternative should be pursued. The following are the model weaknesses that
should be addressed and aimed improved:
Prediction of large leaks
The relative contribution with respect to equipment type
Add specific model for operations involving use of hose(s)
Improve the generic contribution from steel piping
Improve the prediction of the relative contribution from gas and liquid leaks
Improve model prediction of the relative contribution from marginal and significant leaks.
Figure 7.1 UKCS model: Complementary cumulative hole size frequency distribution of observed
leaks at UKCS and model prediction for all equipment types for NCS population dataset (62
installations weighted with individual time in operation in the period 01.01.2001-31.12.2014). The
frequency is obtained by dividing the number of leaks (i.e. observed leaks in HCRD and predicted
for installations at NCS by UKCS model) by the total number of equipment years applicable to leaks
at UKCS and leaks at NCS respectively (all equipment are included, also wells). The number of
equipment years for UKCS installations is 10 939 610 (see TN-3). The number of equipment years
for installations at NCS applicable to this case is 2 878 177 equipment years

Figure 7.2 UKCS model: Normalized complementary cumulative hole size frequency distribution
of observed leaks at UKCS and model prediction for all equipment types for NCS population
dataset (62 installations weighted with individual time in operation in the period 01.01.2001-
31.12.2014). The distribution is obtained by dividing the frequency distribution shown in previous
figure with the frequency for leaks > 1 mm

Figure 7.3 UKCS model: Observed number of leaks 0.1 kg/s versus model prediction for all
installations in the NCS population dataset (181 leaks observed at 62 installations being in
operation in the period 01.01.2001-31.12.2014). The model prediction is 202 leaks

Figure 7.4 UKCS model: Observed number of leaks 0.1 kg/s versus model prediction per
installation for the installations in the NCS population dataset (181 leaks observed at 62
installations being in operation in the period 01.01.2001-31.12.2014). The big blue dot represents
the average for all installations. The dashed red line represents the model target

Figure 7.5 - UKCS model: Observed number of leaks 1 kg/s versus model prediction per
installation for the installations in the NCS population dataset (54 leaks observed at 62 installations
being in operation in the period 01.01.2001-31.12.2014). The big blue dot represents the average
for all installations. The dashed red line represents the model target

Figure 7.6 UKCS model: Observed number of leaks versus model prediction per leak category for
all installations in the NCS population dataset (181 leaks observed at 62 installations being in
operation in the period 01.01.2001-31.12.2014)

Figure 7.7 UKCS model: Observed number of leaks versus model prediction per equipment type
for all installations in the NCS population dataset (181 leaks observed at 62 installations being in
operation in the period 01.01.2001-31.12.2014). The UKCS model does not include models for hoses
used in temporary operations. The contribution from hoses is stemming from flexible piping. The
contribution from storage tanks and vents is not modelled
Figure 7.8 UKCS model: Relative number of observed leaks and model prediction per equipment
type for all installations in the NCS population dataset (181 leaks observed at 62 installations being
in operation in the period 01.01.2001-31.12.2014). The coloured bars represent the individual in-
stallations. The relative distribution for equipment types observed at UKCS is also shown. The UKCS
model does not include models for hoses used in temporary operations. The contribution from
hoses appearing is stemming from flexible piping. The contribution from storage tanks and vents is
not modelled

Figure 7.9 UKCS model: Complementary cumulative leak frequency distribution of observed leaks
at NCS and model prediction per equipment type for all installations in the NCS population dataset
total number of installation years for the NCS population dataset is 802. The contribution from
hoses is stemming from flexible piping
Figure 7.10 UKCS model: Normalized complementary cumulative leak frequency distribution of
observed leaks at NCS and model prediction per equipment type for all installations in the NCS
population dataset (181 leaks observed at 62 installations being in operation in the period
01.01.2001-31.12.2014). The distributions are obtained by dividing the complementary cumulative
leak frequency distributions in previous figure with the total leak frequency (i.e. the total
aggregated normalized frequency equals 1.0). The contribution from hoses is stemming from
flexible piping

Figure 7.11 UKCS model: Observed number of gas leaks versus model prediction per leak
category for all installations in the NCS population dataset (180 leaks observed at 62 installations
being in operation in the period 01.01.2001-31.12.2014)

Figure 7.12 UKCS model: Observed number of liquid leaks versus model prediction per leak

Figure 7.13 - UKCS model: Complementary cumulative distribution of leaks at NCS and model
prediction originating from valves for all installations in the NCS population dataset (85 leaks
observed at 62 installations being in operation in the period 01.01.2001-31.12.2014)
prediction originating from standard flanges for all installations in the NCS population dataset (28
leaks observed at 62 installations being in operation in the period 01.01.2001-31.12.2014)

prediction originating from instruments for all installations in the NCS population dataset (17 leaks
prediction originating from steel piping for all installations in the NCS population dataset (24 leaks

Figure 7.17 UKCS model, Significant leaks: Complementary cumulative leak frequency
distribution of observed leaks at NCS and model prediction per equipment type for all installations
in the NCS population dataset (181 leaks observed at 62 installations being in operation in the
period 01.01.2001-31.12.2014). The frequency is obtained by dividing the number of leaks by the
number of installation years. The total number of installation years for the NCS population dataset
is 802. The contribution from hoses is stemming from flexible piping
Figure 7.18 UKCS model, Marginal leaks: Complementary cumulative leak frequency distribution
of observed leaks at NCS and model prediction per equipment type for all installations in the NCS
01.01.2001-31.12.2014). The frequency is obtained by dividing the number of leaks by the number
of installation years. The total number of installation years for the NCS population dataset is 802.
The contribution from hoses is stemming from flexible piping

8 Parameterisation only based on NCS data
8.1 General
In this section, models developed for hoses, steel pipe and compact flanges based solely on data
gathered from installations on the NCS are presented.
The explanation for the separate methodology for development of leaks from these types of
equipment is:
Hoses
No data available in HCRD on leaks occurring under temporary operations involving use of
hoses. The defined model parameters are therefore solely based on data gathered from
installations on the NCS
Steel pipe
The quality of the population data in HCRD for steel pipes is judged to be poor. Hence, the
model is parameterized based on a subset of the NCS population dataset where equipment
counts of length of steel pipe are available. However, available data in HCRD on equipment
size dependency have been used
Compact flanges
No data is available from UKCS and only limited data available from installations on NCS. A
separate assessment is performed to set the model parameters.
8.2 Hoses
The validation results for hoses (leaks occurring during operations involving the use of hoses) for
the full NCS population dataset is presented in Figure 8.1. The full NCS population dataset should
be used as the total population data (on average 150 hose operations per installation) for all in-
stallations is considered to rather represent all installations on NCS than the subset of 62 install-
ations (denoted NCS population dataset). The population data for hoses are presented in TN-2.
Using 150 hose operations per installation as basis for the estimated leak frequency per hose
operation is judged to be somewhat conservative, i.e. it will result in a higher leak frequency per
hose operation than the actual leak frequency per hose operation at NCS. 150 are used to
account for uncertainty (see TN-2). Moreover, the number of leaks originating from hoses at
installations on the NCS appears to be decreasing (see TN-2) with time in the period 2001-2014.
In total it is believed that the model for hoses is conservative, but the uncertainty is significant.
An improved population dataset of hose operations should be developed to improve the
precision of the leak frequency originating from hoses.
The resulting parameters are presented in Table 8.1.
See TN-5 Appendix A for guidance of how to estimate leak frequency from hoses in QRAs.

Figure 8.1 Complementary cumulative frequency of leaks from hoses versus model prediction for
all installations in the full NCS population dataset (191 leaks in total observed at 74 installations
being in operation in the period 01.01.2001-31.12.2014).) The recorded number of incidents
stemming from hoses is 11
Table 8.1 - Model properties for hose based on NCS population dataset (see TN-2)
() It is judged that there is no dependency with equipment

dimension, which means that () is set to 1.
() Review of the leaks from hoses at NCS indicates that the

fraction of full bore holes should be high. The hole size is
only known in a fraction of the incidents, but based on the
qualitative description in the accident investigation reports it
is found that the hose did rupture in several of the 11 recor-
ded incidents (i.e. rupture of the hose or failure of coupling
to the process system). Another indicator for that the full
bore hole fraction should be high is the high full bore hole
fraction extracted from HCRD for flexible piping for the
period 1992-Q12015 (see Figure 5.6). Although the opera-
tion and use of hoses and flexible piping is different (flexible
piping is permanently in operation), flexible piping can be
considered to an equivalent type of equipment.
No evidence for that () should be dependent on hose
dimension has been identified.
Based on the above, it is concluded to estimate () by
setting to zero and to 0.2. 0.2 means that a full
bore hole (i.e. rupture) will occur in 20 % of the cases.
-4
Significant leak: 110 per operation
-5
Marginal leak: 210 per operation

8.3 Steel pipe
() and () for steel pipe has been estimated based on HCRD (see Figure 5.5 and Figure
5.14). The population data of steel pipes in HCRD is however judged to be unreliable, both in
terms of incidents attributed to steel pipe and the population data. The estimate of for steel
pipe in the UKCS model based on HCRD is therefore not recommended. The failure modes
/causes Opened and Left open is stated for considerable fraction of the incidents in HCRD
attributed to steel piping. This does not necessarily correspond with the properties of the
equipment type steel piping, and it is questioned whether these incidents rather were related to
valves.
for steel pipes (per meter steel pipe) have therefore been estimated by utilizing the available
equipment counts in the NCS population dataset. Equipment counts are available for 24 out of
the 62 installations in the NCS population dataset.
The number of observed leaks from steel pipes in the period 01.01.2001 31.12.2014 is 7 for
those 24 installations. The total number of leaks occurring at the 24 installations in the period is
39. The fraction resulting from steel pipes is hence about 18 %. The average contribution from
steel pipes is 14 % for the entire NCS population dataset. The fraction is sensitive to randomness
in such small population, and the observed steel pipe fraction for the population subset is con-
sidered to be in accordance with the entire NCS population dataset.
Running the validation model for the 24 installations based on the parameters established for the
model denoted NCS model (see Chapter 9.3) results in an estimate of 6.8 leaks originating from
steel pipes using the average pipe fraction of 14 % (see TN-2 and section 9.2.4), which equals
the number of leaks stemming from steel pipes at the 24 installations. Based on this it is
concluded that should be estimated based on targeting 7 observed leaks in the NCS
population dataset subset of 24 installations. This results in the following total for steel pipe:
-5
, = 2.810 per year per meter steel pipe
The resulting is about a factor of two less than the for steel pipes estimated based on
-5
HCRD (, , = 6.310 per year per meter steel pipe).
The fraction marginal leaks is set to 12 % (see section 9.2.4), which result in the following
for split on the leak scenarios (significant and marginal leak):
-5
, , = 2.510 per year per meter steel pipe
-6
, , = 3.410 per year per meter steel pipe
8.4 Compact flange

The following data are available with regard to establish specific model parameters for compact
flanges:
No population data on equipment years covering the use of compact flanges at installations
located on the UKCS is available. It is not known whether any of the leaks associated with
standard flanges in HCRD actually did stem from a compact flange
No leaks in the NCS dataset (seen TN-2) are known to be originating from compact flanges.
Furthermore, it is not known that leaks having an initial leak rate above 0.1 kg/s originating
from compact flanges have occurred on land based facilities in Norway
The total number of equipment years for compact flanges installed in riser systems, subsea
systems and process systems (both offshore and onshore) since the first compact flange was
installed is estimated to be in the interval 500,000 1,000,000 years (Ref. /3/). This figure is
based on interview with vendors supplying compact flanges to industry in Norway.

Based on this, it is concluded that the underlying leak frequency for compact flanges should
comply with 0 leaks occurring in 500,000 equipment years at installations on the NCS. This is
considered to be a conservative approach as the number of equipment years most likely is
significantly larger than 500,000 equipment years.
Based on a brief review of causes for leaks it is judged that the probability for large holes arising
compact flanges are limited. The probability for latent causes, which is important for standard
flanges (see section 5.2.3 and 5.3.3), is significantly less because correct installation can be
visually inspected (a compact flange is only installed correctly if there is no visual opening
between the flange surfaces). Opening of a pressurized bolted joint by mistake have caused some
sudden large releases of fluids from standard offshore (ASME) flanges and clamp connectors. This
event may also be possible with compact flanges, but it is highly unlikely because the design will
provide a pre-warning small leakage far before a critical leakage will occur, and thus enable the
operator to reverse the operation.
The compact flange is also judged not to be vulnerable in an over-pressurization scenario because
when assembled, this bolted joint is normally designed to remain leak tight beyond the yield
stress limit of the adjoining pipe.
Leakage can develop from bolt relaxation due to the dynamics of gasketed bolted joints such as
standard offshore (ASME) flanges when exposed to severe vibrations or other cyclic loads. With
"Non-gasketed" compact flanged joints this event is not possible because the assembled joint
behave static through all load situations within allowable loads for the adjoining piping system.
Analysis and destructive testing have demonstrated that welded joints outside the compact
flanged connection will always fail before leakage can develop in the bolted joint when the
piping system is subject to fatigue loads.
Based on operational experience, the failure mode thought to be the most likely cause for a
future leak in a compact flange is degradation caused by a corrosive internal medium carving out
a small hole in between the flange surfaces. Such a process will develop slowly over time and will
most likely be detected. If a leak finally materializes, the hole is probably quite small, i.e. in the
range of a few millimetres.
Based on the above (supported by Ref. /3/)), the suggested model parameters are presented in
Table 8.2. All leaks from compact flanges are considered to be significant leaks, which is based
on that the most likely failure mode (small hole in between the flange surfaces due to corrosion)
probably would result in a significant leak during normal operation. Figure 8.3 displays model
result for various equipment dimensions. The model for standard flanges is included for
comparison. The plot shows that the deviation between the models increase with increasing hole
size. For big hole sizes, the relative difference between a standard flange and a compact flange
approach a factor of 100. This is because the fraction ruptures (i.e. ()) is judged to be
considerable less for compact flanges opposed to standard flanges.
The stochastic uncertainty of the given data (0 leaks in 500,000 equipment years) can be ass-
essed based on the Poisson distribution. Application of the Poisson distribution for this purpose
requires that the leaks occur independently. This is not entirely true as it is reasonable to assume
that lesson learnt from incidents will have some effect on the likelihood of leaks occurring in the
future.
Figure 8.2 shows the Poisson distribution for 500,000 equipment years based on the standard
-5
flange failure rate in PLOFAM ( = 2.610 per year for significant and marginal leak in total,
which applies for equipment dimension in the interval 3 - 4 for standard flanges (i.e. () 1)).
-6
Using a correction factor of 10 for compact flanges, and rounding up to 3.010 per year, results
in a reasonable likelihood of observing zero leaks in 500,000 equipment years (i.e. about 20 %).
The expected number of leaks using 20 % as the acceptable probability for being fortunate in
500,000 equipment years becomes 1.5. This model accounts for unknown leaks from compact
flanges. Note also that the actual number of compact flange years most likely is significantly
higher than 500,000.

The methodology described above for setting the parameters for compact flanges is stricter than
the approach used when setting the parameters for all other equipment types in PLOFAM. In
general, best estimate is targeted. The rational for the conservative approach is that the
uncertainty associated with the available data is more prominent. It is expected that an even
lower for compact flanges can be justified in the future based on an improved quality of the
data.
Table 8.2 - Model properties for hose based on NCS population dataset (see TN-2)
() It is judged that there is no dependency with equipment

dimension, which means that () is set to 1
() No evidence for that () should be dependent on com-

pact flange dimension has been identified. It is concluded
to estimate () by setting to zero and to 0.001.
0.001 means that a full bore hole (i.e. rupture) will occur
in 0.1 % of the cases
-6
Significant leak: 3.010 per year
Marginal leak: disregarded
Figure 8.2 - Poisson distribution for 500,000 equipment years based on compact flange failure rate
of = 3.010 per year and standard flange failure rate of = 2.610 per year. The calculation
-6 -5
for a standard flange is based on a dimension in the interval 3 - 4

Figure 8.3 PLOFAM model for significant leaks for standard flanges and compact flanges for
various equipment dimensions
9 Best model fit to observed leaks at installations on the

NCS
9.1 General
In order to analyse the difference in underlying frequency distributions for the different
equipment types, a model alternative is developed where it is attempted to obtain the best
possible fit to leaks observed at installations on the NCS (for the period 2001-2014). This is
achieved by running the validation model in an iterative process. This model is denoted the NCS
model.
The resulting model NCS model alternative is presented in this chapter. The model parameters
can be found in Appendix A.
9.2 Model parameters

9.2.1 Hoses
The model for hoses is implemented in accordance with the model developed for installations on
the NCS (see Chapter 8.2). The population data for hoses presented in TN-2 (see Chapter 9.5 in
TN-2) is used.
9.2.2 Producing well and Gas lift well
The number of wells in production in the HCR population data is judged to be more uncertain
than population data of wells established for installations on the NCS. It is therefore concluded to
use the NCS data to adjust the parameters for Producing well and Gas lift well.

The frequency for both Producing well and Gas lift well is adjusted to reproduce the
observed number of leaks for the full NCS population dataset (i.e. the population of 74
installations). In practice this means that the total (i.e. sum of Significant and Marginal leak)
for producing well is adjusted to reproduce the single observed event. The number of leaks from
wells in the NCS population dataset (i.e. 62 installations) is zero. There is not observed any leaks
from wells equipped with gas lift in the NCS dataset. for Gas lift well is set based on that the
relative contribution from Gas lift well on NCS is the same as for installations on the UKCS. This
can be expressed by the following:
,
, =
, ,
9.2.3 Contribution steel pipe

The relative contribution from steel pipes extracted from the HCR database is considered to be
uncertain (see Chapter 10.2.3). In any case, the model should reflect the relative contribution for
leaks at NCS. Hence, the fraction based on HCRD (24 %) is replaced by 14 %, which is the
average fraction of all leaks originating from steel pipes at installations on the NCS (see Table 5 in
Chapter 6 in TN-2). This fraction applies for the total frequency after the contribution from steel
pipe has been added. This means that the total frequency without steel pipe is to be increased by
the factor
1
1.163
(10.14)
9.2.4 Fraction marginal leak

The fraction of total constituting marginal leaks for instruments, flanges, valves and steel
piping is set to 12 % (see Chapter 5 in TN-2). Hence, 88 % of total for these equipment
types is generating a significant leak.
The fraction of the total constituting marginal leaks for Producing well and Gas lift well is
based on the HCRD data.
The contribution from marginal leaks from hoses is as presented in Chapter 8.2.
The fraction generating marginal leak is set to 0 for all other equipment types, which is in
accordance with leaks observed at installations located on the NCS.
9.2.5 Model for instruments, valves and standard flanges
for instruments, valves and standard flanges is adjusted to exactly reproduce the total
observed number of leaks these equipment types. Furthermore, the slope parameters (i.e. (()))
are adjusted according to the complementary cumulative distribution of leaks at NCS for each of
these equipment types. The parameters of () from the UKCS model are kept unchanged.
The imprecision introduced by classification of leaks originating from instruments in the NCS
dataset relative to how the NCS population data for instruments is collected is not adjusted in the
NCS model. This is adjusted for in PLOFAM (see section 10.2.1)
for these types of equipment are adjusted to account for the leaks stemming from vents and
the two leaks with unknown equipment causing the leak.
for actuated valves, manual valves and actuated ESD valves are calculated by adjusting
proportionally with the shift in for valves in the UKCS model to the NCS model (according
to above).

9.2.6 Model for equipment with low number of incidents
For all other equipment types (process vessels (including atmospheric vessel), compressors,
pumps, heat exchangers, filters and pig traps), the slope parameters defined by () and the
trend with equipment dimension (()) from the UKCS model is kept unchanged. The only
adjustment of the UKCS model parameters is an adjustment of Fhist to exactly reproduce the
total observed number of leaks for each of these equipment types.
It is not observed any leaks originating from heat exchangers in the NCS population dataset. In
order to account for randomness, one leak is attributed heat exchangers. This is based on that
the probability for observation of 0 leaks is 50% given that the expected value is 0.69 (see
further description in section 10.2.6). 1 is obtained by rounding upwards to 1 event. To obtain 1
leak altogether from all types of heat exchangers, for each type of heat exchanger are
adjusted proportionally according to the for each of the type of heat exchanger developed
in the UKCS model.
9.3 Validation of the NCS model

The model validation results for the NCS model are shown in Figure 9.1 through Figure 9.18.
The results show that the model is able to reproduce the important aspects of the observed leaks
at installations in the NCS population dataset. The required changes to the parameters in the
UKCS model to obtain the close fit to the NCS data is considered small. Based on this, it is
concluded that the underlying frequency distribution for process leaks at installations on the NCS
is similar to the distribution observed at installations on the UKCS.
The similarity between UKCS and NCS in terms of occurrence of leaks at offshore installations is
confirmed when comparing the complementary cumulative hole size frequency distribution of
observed leaks at UKCS and the corresponding model prediction, which is shown in Figure 9.1.
The difference between the distributions is small. The main difference is that the frequency for
large holes appears to be higher for installations on the UKCS compared to installations on the
NCS. This result is not consistent with the relative complementary cumulative frequency leak rate
distributions for significant leaks shown in Figure 9.2. The data from installations at the UKCS
indicate smaller fraction large leaks in the period after year 2000. It is judged that this only partly
can be explained by different equipment properties (e.g. equipment type and equipment
dimension) and process conditions (e.g. fluid type, pressure, composition and temperature). The
remaining unexplained observed inconsistency is interpreted as being an expression for the
uncertainty associated with the data sets. One important element is the uncertainty related to the
parameters providing the basis for calculation of the initial leak rate in the HCRD (see Chapter 6
in TN-3).
The performance of the NCS model is broadly equivalent with PLOFAM presented in
Chapter 10.3. The main difference is the contribution from the different types of equipment to
the overall leak frequency. The PLOFAM model performance with regard to prediction of large
leaks (> 10 kg/s) is somewhat better. This is mainly because the contribution to the frequency for
large leaks from all equipment types with few associated incidents (process vessels (including
atmospheric vessel), compressors, pumps, heat exchangers, filters and pig traps) is improved in
the PLOFAM model. The validation model results for the NCS model is therefore not discussed
further here. It should however be noted that the performance of the NCS model with regard to
prediction of the relative contribution from liquid and gas leaks is equivalent with the UKCS
model (and the final PLOFAM model). None of the models represent the relative distribution well.
This is further discussed in Chapter 10.3.

Figure 9.1 NCS model: Complementary cumulative hole size frequency distribution of observed
leaks at UKCS and model prediction for all equipment types for the NCS population dataset (62
for installations at NCS by UKCS model) by the total number of equipment years applicable to leaks
at UKCS and leaks at NCS respectively (all equipment are included, also wells). The number of
equipment years for UKCS installations is 10 939 610 (see TN-3). The number of equipment years
for installations at NCS applicable to this case is 2 998 477 equipment years (the number of
equipment years includes hose operations (150 per installation per year) and the number wells in
production)

Figure 9.2 - Relative complementary cumulative frequency leak rate distributions for significant
leaks for installations on the UKCS for the period 1992-Q1 2015, for UKCS for the period 2001-Q1
2015 and NCS for the period 2001-2014. Note that incidents are extracted with respect to

Figure 9.3 NCS model: Observed number of leaks 0.1 kg/s versus model prediction for all

Figure 9.4 NCS model: Observed number of leaks 0.1 kg/s versus model prediction per

Figure 9.5 - NCS model: Observed number of leaks 1 kg/s versus model prediction per installation
for the installations in the NCS population dataset (54 leaks observed at 62 installations being in
operation in the period 01.01.2001-31.12.2014). The big blue dot represents the average for all
installations. The dashed red line represents the model target

Figure 9.6 NCS model: Observed number of leaks versus model prediction per leak category for

Figure 9.7 NCS model: Observed number of leaks versus model prediction per equipment type for
operation in the period 01.01.2001-31.12.2014). The contribution from storage tanks and vents is
not modelled
Figure 9.8 NCS model: Relative number of observed leaks and model prediction per equipment
stallations. The relative distribution for equipment types observed at UKCS is also shown. The

Figure 9.9 NCS model: Complementary cumulative leak frequency distribution of observed leaks
total number of installation years for the NCS population dataset is 802
Figure 9.10 NCS model: Normalized complementary cumulative leak frequency distribution of
aggregated normalized frequency equals 1.0)

Figure 9.11 NCS model: Observed number of gas leaks versus model prediction per leak category

Figure 9.12 NCS model: Observed number of liquid leaks versus model prediction per leak

Figure 9.13 - NCS model: Complementary cumulative distribution of leaks at NCS and model


Figure 9.17 NCS model, Significant leaks: Complementary cumulative leak frequency distribution
of installation years. The total number of installation years for the NCS population dataset is 802
Figure 9.18 NCS model, Marginal leaks: Complementary cumulative leak frequency distribution of

10 PLOFAM
10.1 General
The final model alternative is presented in this chapter. The recommended model alternative is
developed based on a methodology that weighs the two available datasets.
The final model is denoted PLOFAM.
The parameters are enclosed in Appendix A.
10.2 Model parameters

10.2.1 Hoses
The model for hoses is implemented in accordance with the model developed for installations on
the NCS (see Chapter 8.2). The population data for hoses presented in TN-2 is used (i.e. 150 hose
operations per year per installation).
10.2.2 Producing well and Gas lift well
The number of wells in production in the HCR population data is judged to be more uncertain
than population data of wells established for installations on the NCS. It is therefore concluded to
use the NCS data to adjust the parameters for Producing well and Gas lift well.
The frequency for both Producing well and Gas lift well is adjusted to reproduce the
observed number of leaks for the full NCS population dataset (i.e. the population of 74
installations). In practice this means that the total (i.e. sum of Significant and Marginal leak)
for producing well is adjusted to reproduce the single observed event. The number of leaks from
wells in the NCS population dataset (i.e. 62 installations) is zero. There is not observed any leaks
from wells equipped with gas lift in the NCS dataset. for Gas lift well is set based on that the
relative contribution from Gas lift well on NCS is the same as for installations on the UKCS. This
can be expressed by the following:
,
, = ,
,
10.2.3 Contribution steel pipe

The relative contribution from steel pipes extracted from the HCR database is considered to be
somewhat uncertain. The failure modes /causes Opened and Left open is stated for
considerable fraction of the incidents in HCRD attributed to steel piping. This does not necessarily
correspond with the properties of the equipment type steel piping, and it is questioned whether
these incidents rather were related to valves.
Hence, the fraction steel pipe estimated based on the NCS data is considered more reliable than
the fraction based on HCRD (24 %). The average fraction of all leaks originating from steel pipes
at installations on the NCS is therefore recommended in PLOFAM. The fraction is 14%. This
fraction applies for the total frequency after the contribution from steel pipe has been added.
This means that the total frequency without steel pipe is to be increased by the factor
1
1.163
(10.14)
10.2.4 Fraction marginal leak

The fraction of total constituting marginal leaks for instruments, flanges, valves and steel
piping is set to 12 % (see Chapter 5 in TN-2). Hence, 88 % of total for these equipment
types is generating a significant leak.
The fraction of the total constituting marginal leaks for Producing well and Gas lift well is
based on the HCRD data.

The contribution from marginal leaks from hoses is as presented in Chapter 8.2.
The fraction generating marginal leak is set to 0 for all other equipment types, which is in
accordance with leaks observed at installations located on the NCS.
The marginal leak scenario is considered relevant for all hole sizes. The relative complementary
cumulative frequency leak rate distributions for installations on the UKCS for the period 1992-Q1
2015 (shown in Figure 5.9), demonstrates that the marginal leak scenario has produced initial
leak rates beyond 100 kg/s. The maximum initial leak rate for a marginal leak occurring at an
installation on the NCS in the period 2001-2014 is about 10 kg/s (see Figure 10.1), which is
similar to the maximum initial leak rate observed at installations on the UKCS in the period after
year 2000 (see Figure 5.8). The duration of a marginal leak having a leak rate beyond 100 kg/s is
short (< 1 second), but is not considered unreasonable. For example, sudden release of a small
volume of high pressured gas trapped inside a piece of equipment may represent such a scenario.
A concrete example is opening of a drain plug in the body of a large valve containing gas at high
pressure. Similar incidents have been recorded at installations on the NCS.
on the NCS for the period 2001-2014
10.2.5 Model for valves, standard flanges and instruments

A leak stemming from a valve or a flange associated with instrument connections is in general
recorded as valve or flange leak in the NCS dataset, and not as a leak from the instrument
connection. The equipment counts gathered from installations on NCS are based on a
methodology where valves and flanges associated with the instrument are represented by the
instrument, which is based on the methodology used to establish the population dataset in HCR.
This means that the classification of leaks in terms of type of equipment and the corresponding
NCS population dataset is not consistent.

The equipment count methodology used by the various consultancies providing the equipment
counts vary somewhat. Furthermore, it is expected that leaks originating from a flange or a valve
associated with an instrument connection are in some cases classified as leaks stemming from the
instrument. This means that it is not straight forward to adjust for the inconsistency between the
way leaks stemming from instruments have been in classified in the NCS dataset and the
population data.
There is also uncertainty related to HCR database in this regard. The failure modes /causes
Opened and Left open is stated for considerable fraction of the incidents in HCRD attributed
to steel piping and flanges. This does not correspond very well with the equipment type steel
piping, but is probably more relevant for flanges (may for instance indicate latent errors) and it is
questioned whether some of these incidents rather where related to valves. If so, the relative
contribution fromo these types of equipment for installations on UKCS becomes more similar to
the distribution with respect to equipment types observed at installations on the NCS.
To account for the inconsistency described above, the following adjustment is implemented
(based on an overall evaluation of the data):
for instruments is increased to generate 11 more leaks than in the NCS dataset (17 in
NCS dataset vs. 28 predicted leaks by PLOFAM)
for flanges is reduced to generate 5 less leaks than in the NCS dataset (28 in NCS
dataset vs. 23 predicted leaks by PLOFAM)
for valves is reduced to generate 12 less leaks than in the NCS dataset (85 in NCS
dataset vs. 73 predicted leaks by PLOFAM).
The total targeted number of leaks originating from valves, standard flanges and valves then
becomes 124, whilst the total number of observed leaks from these types of equipment is 130.
This is because the adjustment also needs to account for the targeted overprediction of leaks
originating from other equipment types (see following chapter) and that the overall target for
leaks is 181. Hence, 6 leaks are shifted to other components than valves, flanges and instruments
to match the overall target.
The slope parameters of the hole size distribution for valves, standard flanges and instruments
established based on the UKCS data is kept unchanged. The validation of the NCS model
demonstrated that a slightly better fit is possible by adjusting the slope parameters for these
types of equipment, but the difference is marginal. Hence, it is concluded that the slope
parameters for valves and instruments are concluded to be applicable also for equipment at NCS.
() for valves, flanges and instruments based on UKCS data is also kept unchanged.
This means that it is concluded that the shape of the underlying hole size distribution for valves,
standard flanges and instruments at installations on UKCS and NCS are identical.
for actuated valves, manual valves and actuated ESD valves are taken from the NCS model.
10.2.6 Model for equipment with few incidents
The number of leaks in both datasets stemming from valves, instruments, flanges and steel
piping is considerable. For all other equipment, the number of registered leaks is quite few, and
the parameterisation is sensitive to randomness.
For these equipment types a methodology has been developed that is based on both datasets.
Leaks from UKCS and NCS and the validation methodology are used as basis for tuning the
model parameters. A target is set for each type of equipment for how many leaks the validation
model (LeakPy) should predict for the considered time period. Two targets are set for the number
of leaks with an initial leak rate exceeding 0.1 kg/s and 10 kg/s respectively.
Due to the limitation in the available data, there is no simple rule for how to set these targets.
The data is structured in three data sets summarized in Table 10.1.

Table 10.1 - The three data sets with naming convention
Marginal Significant leaks (quantity > 10 kg and rate > 0.1 kg/s)
Population
(equipment
quantity < 10 kg years)
Small: 0.1-1 kg/s Medium: 1-10 kg/s Large > 10 kg/s
rate >0.1 kg/s
UKCS 1992-
MargUK1 SmallUK1 MediumUK1 LargeUK1 PopUK1
2000
UKCS 2001-
MargUK2 SmallUK2 MediumUK2 LargeUK2 PopUK2
2015
NCS 2001-2014 MargN SmallN MediumN LargeN PopN
Total MargSUM SmallSUM MediumSUM LargeSUM PopSUM
A frequency (f) for significant leaks applicable for the NCS is quantified for each equipment type,
reflecting the number of incidents and the corresponding population. Then the target number of
incidents for the relevant equipment type (for the relevant time period), T, (for LeakPy) is then
calculated as T = f PopN.
The number of significant leaks are denoted SignificantN, SignificantUK2 etc. For example,
SmallN + MediumN + LargeN = SignificantN
In the proposed rule set, a minimum acceptable number of incidents is proposed as basis for the
leak frequency estimation in order to limit the stochastic uncertainty. The proposed minimum
number of incidents used as basis is proposed set to 5. The following cases are used:

Case 1: SignificantN 5: =

+ 2
Case 2: SignificantN < 5 and SignificantN + SignificantUK2 5: =
+2
+ 2+ 1
Case 3: SignificantN + SignificantUK2 < 5: =

An additional rule (relevant only for case 2 and 3) that defines a maximum value for f as follows
is applied: The probability for observing more than SignificantN incidents should not exceed
75 % given f and PopN. The value of 75 % account for uncertainty due to randomness, i.e. we
expect that the actual underlying leak frequency is higher than what is observed. In other words,
it is accounted for that randomness has resulted in an observed frequency that is less than the
true underlying frequency. This approach ensures that randomness related to the total number of
leaks is reflected in the model. Best estimate is used to set the target for leaks > 10 kg/s to avoid
bias in the model with regard to large leaks. Large leaks are rather controlled by the model for
the equipment types that dominates the total frequency distribution.
The Gamma probability distribution (the conjugate distribution for a Poisson process) is applied:
= . (0.75, , )
where
= + 1
1
=

For example, for the tube and shell heat exchangers we have: SignificantN = 0, PopN = 6048
-4 -4
fmax = Gamma.inv(0.75,1,1.6510 ) = 2.2910
0.75 is proposed as a reasonable upper value. This probability is only applicable in cases where
the data from the UKCS indicates a higher frequency than for the NCS, and justify a reduced
frequency to reflect that NCS data are lower than HCRD data.

The same rule is applied to fence the frequency at the lower end. In this case the probability is set
to 0.25 (1-0.75). This does not have any effect in practice for the observed data in our case
(except marginally for Filters).
Using specific rules such as the one proposed will not reflect all aspects of the uncertainty in the
data, such as uncertainty and inaccuracies in population and incident data. Strict use of such
rules will therefore not be the best approach in all cases.
The resulting targets are summarized Appendix C.
It has been concluded to use an amalgamated model for leaks at shell side and tube side for shell
and tube heat exchangers. Separate targets for shell side and tube side have been investigated,
and found to be similar. Thus in the validation process, only parameters for a shell and tube heat
exchanger are developed as part of PLOFAM.
10.3 Validation of PLOFAM

The PLOFAM validation results are shown in Figure 10.2 through Figure 10.17.
The results show that PLOFAM is able to reproduce the important aspects of the observed leaks
at installations in the NCS population dataset. The required changes to UKCS model to obtain the
close fit to the NCS data is considered marginal. Based on this, it is concluded that the underlying
frequency distribution for process leaks at installations on the NCS is similar to the distribution
observed at installations on the UKCS.
This is underlined when comparing the complementary cumulative hole size frequency
distribution of observed leaks at UKCS and the corresponding model prediction, which is shown
in Figure 10.3. The difference is smaller than expected. The main difference is that the frequency
for large holes appears to be higher for installations on the UKCS compared to installations on
the NCS. The small difference is not entirely consistent relative complementary cumulative
frequency leak rate distributions for significant leaks shown in Figure 9.2. This is discussed further
in Chapter 9.3, where it is concluded that observed inconsistency is an expression for the
uncertainty associated with both of the available data sets. The difference in the total cumulative
frequency is due to that PLOFAM has accounted for the uncertainty related to incidents with a
hole size around 1 mm in HCRD through the introduction of the parameter denoted 1 in the
validation process. The difference in the total accumulated frequency between the distribution
for installations on the NC and the observed distribution extracted from HCRD equals the average
value of 1 displayed in Figure 5.15 (i.e. about 1.25).
The main outstanding shortcoming of PLOFAM is the capability to reflect the relative contribution
from liquid and gas leaks. Figure 10.12 and Figure 10.13 shows that the model is not able to
predict the individual contribution from the gas and liquid leaks very well. The model overpredicts
the contribution from liquid leaks considerably. The model predicts the contribution from gas
leaks having a leak rate above 10 kg/s accurately, but underestimates the total contribution from
gas leaks with an initial leak rate less than 10 kg/s. The deviation between the model prediction
and the observed number of leaks in the interval 1 10 kg/s is about 15 %. For leaks less than 1
kg/s, the discrepancy is about 30 %. Looking at the total result (see Figure 10.7), the excessive
contribution from liquid leaks compensate for the under prediction of small gas leaks. Thus the
total model prediction balances the observed number of leaks in the population.
The reason for the difference in model prediction with respect to fluid phase has not been fully
understood. One reason could be imprecise modelling of the leak rate, which particularly is
judged to be important for multi-phase leaks where estimation of the correct density is
challenging. Uncertainty when classifying leaks in terms of fluid phase is also prominent.
Classification in the accident investigation report of the incident is partly based on observed
properties of the release dispersing in the surroundings. Hence, a multi-phase or liquid leaks that
evaporates when expanding to atmospheric conditions may in general be classified as gas leaks.

It may however be the case that the underlying hole size frequency distribution is different for
equipment containing liquid and gas. For instance, the degradation mechanisms could be
different.
An improved model prediction could have been obtained by defining separate model parameters
for equipment containing gas and liquid. As no apparent explanation has been found it has not
been attempted to establish specific models for equipment containing gas and equipment. It is
recommended that this is addressed in future projects.
The time trend in the NCS population data is analysed in TN-2 (see Chapter 10), where it is
concluded that there is a significant trend with time for all leaks. The average leak frequency per
component in the NCS population datasets is shown in Figure 10.2 for the entire period
(01.01.2001 31.12.2014) and for the recent 5 year period. The average frequency for the last 5
years of the period is about 40 % less than the average for the entire period. The future leak
frequency for installations on the NCS is expected to be more in line with the average for the
recent 5 years. Note that for the period after 2007, PLOFAM overpredict the frequency for both
gas and liquid leaks over the entire spectre of initial leak rates.
The 30 % underprediction in frequency for small gas leaks by PLOFAM balance the difference in
average historical leak frequency for the entire period and the recent 5 year period. It is therefore
concluded that PLOFAM is able to predict future small gas leaks at installations on the NCS with
adequate precision. All other leaks, i.e. all liquid leaks and gas leaks having an initial leak rate
larger than 1 kg/s will most likely be overpredicted by PLOFAM.
In Figure 10.18 and Figure 10.19, the complementary cumulative leak frequency distribution of
observed leaks per scenario type (i.e. significant leaks and marginal leaks) is shown. The result
demonstrates that PLOFAM is able to reproduce the contribution from the two leak scenario
types.
The uncertainty related to the PLOFAM is discussed further in Chapter 11. The model prediction
per installation is discussed (i.e. the variance displayed Figure 10.5) as well as the sensitivity with
respect to uncertain aspects related to the data the model is based upon.

Figure 10.2 - Average frequency for leaks ( 0.1 kg/s) per component for various time periods and
the two NCS population datasets (see Chapter 11 in TN-2 for description of the results)

Figure 10.3 PLOFAM: Complementary cumulative hole size frequency distribution of observed
leaks at UKCS and model prediction for all equipment types for the NCS population dataset (62
for installations at NCS by the PLOFAM model) by the total number of equipment years applicable
to leaks at UKCS and leaks at NCS respectively (all equipment are included, also wells). The number
of equipment years for UKCS installations is 10 939 610 (see TN-3). The number of equipment years
for installations at NCS applicable in this case is 2 998 477 equipment years (the number of
equipment years includes hose operations (150 per installation per year) and the number wells in
production)

Figure 10.4 PLOFAM: Observed number of leaks 0.1 kg/s versus model prediction for all
operation in the period 01.01.2001-31.12.2014). The model prediction is 180.6 leaks

Figure 10.5 PLOFAM: Observed number of leaks 0.1 kg/s versus model prediction per

Figure 10.6 PLOFAM: Observed number of leaks 1 kg/s versus model prediction per installation

Figure 10.7 PLOFAM: Observed number of leaks versus model prediction per leak category for all

Figure 10.8 PLOFAM: Observed number of leaks versus model prediction per equipment type for
operation in the period 01.01.2001-31.12.2014). The contribution from storage tanks and vents is
not modelled
Figure 10.9 PLOFAM: Relative number of observed leaks and model prediction per equipment

Figure 10.10 PLOFAM: Complementary cumulative leak frequency distribution of observed leaks
Figure 10.11 PLOFAM: Normalized complementary cumulative leak frequency distribution of

aggregated normalized frequency equals 1.0)

Figure 10.12 PLOFAM: Observed number of gas leaks versus model prediction per leak category

Figure 10.13 PLOFAM- Observed number of liquid leaks versus model prediction per leak

Figure 10.14 PLOFAM: Complementary cumulative distribution of leaks at NCS and model
Figure 10.15 PLOFAM: Complementary cumulative distribution of leaks at NCS and model

Figure 10.16 - PLOFAM: Complementary cumulative distribution of leaks at NCS and model
Figure 10.17 - PLOFAM: Complementary cumulative distribution of leaks at NCS and model

Figure 10.18 - PLOFAM: Significant leaks: Complementary cumulative leak frequency distribution
Figure 10.19 - PLOFAM: Marginal leaks; Complementary cumulative leak frequency distribution of

11 Robustness PLOFAM
11.1 Introduction
In this chapter, the uncertainty related to the capability of the model to predict the occurrence of
future leaks at installations on the NCS is discussed.
The following aspects are discussed in separate sections:
General evaluation of quality of data basis
The variance in prediction of installation specific leak frequency
Quality of the NCS population dataset
Effect of trend with time and the prediction of future gas leaks
Randomness related to the observation of incidents.
11.2 General evaluation of quality of data basis

The quality and limitations of the data used as basis for the parameterisation of the PLOFAM
parameters is fundamental for the precision of the model. The quality of the data basis is
discussed throughout the report, both in the technical notes presenting the data basis (TN-2 and
TN-3), but also in discussion of the results generated by the validation model.
It is judged that the elements affecting the quality and limitations of the data are understood, but
some of them may be hard to quantify. On a high level, the frequency distributions based on
data extracted from the HCR database and the NCS database is similar. This is considered to be a
strong argument for that the PLOFAM model is based on a solid understanding of the data basis.
This can for instance be seen from Figure 10.3, where it is shown that the average
complementary cumulative leak frequency distribution for (1) all equipment types at UKCS and
(2) modelled by PLOFAM based on the NCS population dataset is similar.
The main difference between the data from UKCS and NCS installations are the relative
contribution from different types of equipment. This may be due to randomness, in particular for
equipment with a low number of equipment years, but also due to differences in the way leaks
are logged in HCRD opposed to in the NCS dataset. The sensitivity due to a low number of leaks
for certain types of equipment is handled by a separate methodology described in section 10.2.6
in this technical note. The difference in classification of incidents in the HCRD and NCS datasets
in terms of equipment type has been evaluated and incorporated in the model (see section
10.2.5 in this technical note).The uncertainty related to parameterisation of the model for hoses,
steel pipes and compact flanges are discussed in Chapter 8 in this technical note.
Important factors affecting the quality of the NCS population dataset are discussed separately in
section 11.4.1 and section 11.4.2 below (a review of the quality of the NCS data is also included
in TN-2).
The validation process has shown that the main outstanding shortcoming of PLOFAM is the
capability to reflect the relative contribution from liquid and gas leaks, which is briefly discussed
in section 10.3 in this technical note (see Figure 10.12 and Figure 10.13). The causes are not fully
understood, but it is believed that classification of leaks at NCS installations in terms of fluid
phase is important. Classification in the accident investigation report of the incident is partly
based on observed properties of the release dispersing in the surroundings. Hence, a multi-phase
or liquid leaks that evaporates when expanding to atmospheric conditions may in general be
classified as gas leaks. The uncertainty related to prediction of the distribution with respect to
fluid phase is accounted for by using the period 2001-2014 for parameterisation of the PLOFAM
parameters. The observed leak frequency in the period 2001-2014 is considerably higher than in
the later part of the period (see section 11.5 below). Suggested future work on this topic is
described in Chapter 15.

11.3 Prediction of installation specific leak frequency
In Figure 10.5, the observed number of leaks 0.1 kg/s versus model prediction per installation
for the installations in the NCS population dataset is shown for PLOFAM. The model will on
average predict the number of observed leaks on the installations in the NCS population dataset.
The results for the various installations are spread around the expected value and the average
number of leaks for all installations are perfectly predicted (the big blue dot). However, the
variance is considerable.
A perfect model would produce a prediction of number leaks that is located on a line where the
number observed leaks equals the predicted number of leaks (the red dashed line in Figure 10.5).
The general explanation for the observed variance is deficiencies related to the modelling
principle and randomness.
Analysis of the causes for leaks (e.g. Ref. /2/) has demonstrated that the properties of the
equipment and the process conditions are far from the only factors affecting the occurrence of
leaks. Human and organisational factors have to be taken into the picture to fully understand the
underlying drivers for leaks. The correlation between such factors and the occurrence of leaks are
very complex, and PLOFAM does not attempt to model these casual factors. PLOFAM is based on
a modelling approach where the leak frequency is estimated based on equipment counts, process
conditions and an averaged frequency distribution for the occurrence of holes per standard
equipment types. Given this modelling approach and the complexity of human and
organisational factors, the observed variance per installation seen in Figure 10.5 is considered
reasonable. Moreover, the averaged frequency distribution for the occurrence of holes per
standard equipment types does not strictly apply for the equipment at the individual installation;
it applies for the average of the population. It is reasonable to argue that there is technical (e.g.
design, material properties) differences between the same standard equipment at different
installations (i.e. a 6 inch centrifugal pump operating produced oil at low pressure and delivered
by one vendor may be somewhat different in terms of the underlying frequency for leaks relative
to a 2 inch centrifugal pump operating condensate at high pressure at another installation). The
point being made here is that the causes for leaks are very complex and difficult to predict. In
short, this can be summarized saying that the frequency for leaks originating from a piece of
equipment is not an inherent property of the equipment.
PLOFAM is targeting the average leak frequency for the entire population for the entire period
2001-2014, and a variance around the average value is expected. This means that it is a target
that PLOFAM predict the trend with regard to size of the installation and the number of years in
operation. In other words, PLOFAM should predict the average for installations with few recorded
incidents equally good as installations where a high number of leaks have occurred.
Looking more closely at the results in Figure 10.5, it is clear that the model tends to underpredict
the observed number of leaks for installations that has experienced a high number of leaks, and
overpredict the number of leaks for installations where few leaks have been observed in the
period (01.01.2014 31.12.2014). This indicates that there is an undesired trend in the dataset
that is not explained by PLOFAM.
Figure 11.1 displays the validation results with LeakPy for PLOFAM for the 8 year period starting
at 01.01.2007 and ending at 31.12.2014. The results show that the deviation from the targeted
average is less prominent. Still there is quite a few installations with a low number of observed
leaks (0-1) where PLOFAM overpredict the number of leaks. The variance around the average
value for installations with many leaks (2-4 leaks) appears to be symmetrical. A fraction of the
apparent overprediction for installations with few incidents can be explained by randomness.
More important for the result is however the fact that the PLOFAM parameter values (i.e. )
has not been adjusted for the reduced observed frequency of leaks in the period after 2006. The
parameters used are still based on the entire period.

The improved performance for installations with many leaks for the shorter time period is explai-
ned by the decreasing trend in number of observed leaks with time throughout the period 2001-
2014. The number of leaks has decreased with time for all installations in the dataset. The dec-
rease is however most prominent for the installations being in production the entire period.
These installations tend also to be somewhat larger than then the newer installations. This is the
installations where the greatest number of leaks is observed, and is the installations the model
tends to underpredict. The decreasing trend with time with respect to observed leaks is most
prominent in the period 2001-2006 (see Chapter 10 in TN-2). The model performance per
installation is therefore better for the time period after 2006. But it is known that there also has
been a decreasing trend with time after 2006 for installations at NCS, which also affects the
prediction for installations with few leaks in the period after 2006 (observed 75 leaks at
installations on the NCS while 106 is predicted by LeakPy in the same period). This means that it
can be expected that PLOFAM will on average overpredict the future frequency for leaks at
installations on the NCS.
The variance around the average value is also investigated by calculating how many of the obser-
vations (one installation is an observation in this context) that can be considered to be extreme.
This investigated by applying the Poisson distribution and assuming the PLOFAM prediction
represents the underlying leak frequency per installation. An extreme event is set to be an event
it is beyond the number of leaks defined by a 5 % confidence interval (the aggregated probability
beyond the number of leaks defined by the confidence interval is 2.5 % at each side of the mean
of the distribution). For PLOFAM for the period after 2006, only 1 installation is rejected in such a
case (see Table 11.1). 1 installation equals less than 5 % of the number of installations (5 % out
of 62 is 3). This is interpreted as an indicator for that the observed variance is reasonable.
Based on above, it is in total concluded that PLOFAM capability to predict the installation specific
leak frequency is acceptable. Here it must be emphasized that the PLOFAM is equally valid for an
installation where PLOFAM underpredict the observed number of leaks as installations where
PLOFAM overpredict the observed number of leaks. PLOFAM target the average for all
installations located on the NCS.

Figure 11.1 PLOFAM: Observed number of leaks 0.1 kg/s versus model prediction per
installation for the installations in the NCS population dataset (75 leaks observed at 62 installations
being in operation in the period 01.01.2007-31.12.2014). The big blue dot represents the average
for all installations. The dashed red line represents the model target

Table 11.1 Poisson process analysis for PLOFAM prediction for leaks at NCS in the period 2007-
2014
Installation PLOFAM Historical Poisson distribution Outside 5%
mean P observed P observed confidence
events given events given interval
PLOFAM mean PLOFAM mean
Platform 58 0.47 0 0.62 0.38 Accepted

Installation PLOFAM Historical Poisson distribution Outside 5%
mean P observed P observed confidence
events given events given interval
PLOFAM mean PLOFAM mean
Platform 63 1.42 4 0.98 0.02 Rejected
11.4 Quality of the NCS population dataset

11.4.1 Sensitivity studies with validation model
In TN-2, it is stated that the general quality of the NCS populations dataset is high (see Chapter
9.8 in TN-2). In order to assess the sensitivity to bias in the population dataset, a couple of
sensitivity studies have been performed with the validation model (LeakPy).
The following two cases have been run:
PLOFAM for the full NCS population dataset. This dataset consist of 74 installations and
capture all observed leaks in the period 2001-2014. The result from this sensitivity is used to
assess the uncertainty related to equipment counts. It is argued in TN-2 that the number of
equipment counts related to the 12 additional installations in the full NCS population
outweigh the excessive number of equipment counts related to modifications to installations
in the NCS population dataset in the period. The latter means that the number of equipment
years for installations subjected to major modifications may be exaggerated because the
installed equipment is assumed to be in operation the entire time period (2001-2014) in the
validation model. This may lead to underprediction of by use of the validation model

PLOFAM for the full NCS population dataset and with 15 % lower pressure (and density for
gas leaks) than specified in the NCS population dataset. This is investigating the effect of
systematic bias in terms of pressure in all QRAs of which the equipment counts are gathered
from. The concern is that the pressure at installations being in production in the entire period
has probably dropped throughout the period. Hence, if an upper estimate of pressure has
been used as basis for the QRA and is assumed to apply for the entire period in the validation
model, the validation model may indicate overprediction, particularly for large leaks (> 10
kg/s), when in fact the situation is the other way around. In general, it is believed that the
effect of pressure is small. For some recently updated QRAs used as basis for the NCS
population dataset, the pressure will in fact be lower than the average pressure applicable for
the entire period.
The resulting complementary cumulative leak frequency distribution of observed leaks at NCS
and the corresponding model prediction is shown in Figure 11.3 and Figure 11.4. The prediction
of total number of leaks for the case with unaltered pressure is shown in Figure 11.2.
The results for the case with unchanged pressure (Figure 11.2 and Figure 11.3) shows that
PLOFAM will overpredict the total number of leaks as well as the number of leaks larger than 10
kg/s. This means that some effect related to excessive number of equipment counts are
accounted for in the validation model for the full NCS population dataset (i.e. all 74 installations)
in the period 2001-2014. The total effect on the number of equipment years due to
modifications to installations in the period is however judged to be quite small. In TN-2 it is
argued that the effect on the number of equipment years is in order of magnitude 1 %. 5% is
judged to be a upper estimate, which still would not affect the conclusions being made. It is
however recommended that future projects do focus on further enhancement of the quality of
the NCS population dataset.
The sensitivity with regard to pressure, shown in Figure 11.4, demonstrates that a moderate bias
in terms of pressure will not affect the validation results much. The main effect is on very large
leaks. The robustness of the model in terms of large leaks is however accounted for by the trend
with time, which is discussed in the following chapter.

Figure 11.2 PLOFAM: Observed number of leaks 0.1 kg/s versus model prediction for all
installations in the full NCS population dataset (190 leaks observed at 72 installations being in
Figure 11.3 - PLOFAM: Significant leaks: Complementary cumulative leak frequency distribution of

Figure 11.4 - PLOFAM: Significant leaks, reduced pressure by 15% for all process segments in the
full population dataset : Complementary cumulative leak frequency distribution of observed leaks
11.4.2 Adjustment for operational time in equipment counts

Chapter 9.8 in TN-2 brings attention to uncertainty related to whether adjustment for time in
operation has been included in the equipment count database. This may affect all equipment
types, but it is judged that the operational time is most important for pig traps, which is typically
pressurized only a fraction of time.
A significant fraction of the equipment counts for pig traps is registered with a quantity less than
1, which indicates that the time in operation is less than 100%. To investigate the effect of not
adjusting for time in operation, the number of pig trap years has been calculating setting all
counts with a quantity less than 1 to 1. This is considered to be a reasonable approximation as it
is expected that there is only one pig trap per ESD segment. This results in an increased number
of pig trap years of about 10% (indicating that adjustment has been performed for new
installations). This adjusted equipment count for pig traps has been used to update the target
values for equipment types with a small number of recorded incidents as described in Chapter
10.2.6. The resulting target for the total frequency is increased by a factor of about 5%. This is
considered insignificant for the model parameters for pig traps.
The effect on other equipment types is believed to be considerably less. Hence, the effect of the
uncertainty related to adjustment of operational time in the population dataset is considered
marginal. Anyway, this should preferably be reviewed and corrected in future projects updating
the population database.
Note that the equipment counts for calculation of frequencies based on PLOFAM should not be
adjusted for operational time (see Appendix A TN-5). This is in line with that leaks often occur in
connection with maintenance, modification or start up/shut down (e.g. latent failure modes such
as a valve normally closed left in open position). The actual operational time will not be important
for the occurrence of such leaks.

11.5 Time trend and prediction of gas leaks
The time trend in the NCS population data is analysed in TN-2 (see Chapter 10), where it is con-
cluded that there is a significant trend with time for all leaks. The average frequency for the last 5
years of the period is about 40 % less than the average for the entire period. The future leak
frequency for installations on the NCS is expected to be in line with the average for the recent 5
years.
In order to check the effect of the trend in time in detail, the validation model has been run for
the period 01.01.2007 to 31.12.2014. To further test the robustness, one additional large gas
leak (initial leak rate of 100 kg/s) is added to check whether the model account for such an
incident in the near future. The additional event can also be interpreted as an event representing
the uncertainty related to initial leak rate of the leaks that has occurred.
The trend analysis presented in TN-2 demonstrates that there also is an underlying trend in the
period 2007 onwards. The floating average shown in Figure 11.9 for the installations in the NCS
population datasets demonstrates that there is constant decreasing trend for the entire period.
The resulting complementary cumulative leak frequency distribution of observed leaks at NCS
and the corresponding model prediction is shown in Figure 11.5 and Figure 11.7 for total (signi-
ficant + marginal leaks) and significant leaks respectively. The results demonstrate that PLOFAM
will overpredict the frequency for all leak rates considerably in the period after 2007. This follows
from the fact that the PLOFAM parameters have been set based on the entire period 2001-2014.
The historical frequency for the period 2001-2014 is considerably higher than for the period
2007 2014.
Figure 11.6 and Figure 11.8, that applies for gas leaks, demonstrates that the difference between
the observed number of leaks and the prediction is small. This result demonstrates that the effect
of the trend with time will account for PLOFAM deficiency in terms of prediction of the individual
contribution from gas and liquid leaks. Based on this result, it is concluded that PLOFAM model is
expected to predict the future occurrence of gas leaks at installations on the NCS. The prediction
of liquid leaks is expected to be excessive. In total the model will generate somewhat
conservative estimates of the frequency for loss of containment from process systems at
installations on the NCS.

Figure 11.5 - PLOFAM: Complementary cumulative leak frequency distribution of observed gas and
liquid leaks at NCS and model prediction per equipment type for all installations in the NCS
of installation years. The total number of installation years for the NCS population dataset in the
period is 471
Figure 11.6 - PLOFAM: Complementary cumulative leak frequency distribution of observed gas
leaks at NCS and model prediction per equipment type for all installations in the NCS population
dataset (54 leaks observed at 62 installations being in operation in the period 01.01.2007-
31.12.2014). The frequency is obtained by dividing the number of leaks by the number of
installation years. The total number of installation years for the NCS population dataset in the
period is 471

Figure 11.7 PLOFAM (significant leaks): Complementary cumulative leak frequency distribution of
observed gas and liquid leaks at NCS and model prediction per equipment type for all installations
in the NCS population dataset (67 leaks observed at 62 installations being in operation in the
period 01.01.2007-31.12.2014). The frequency is obtained by dividing the number of leaks by the
number of installation years. The total number of installation years for the NCS population dataset
in the period is 471
Figure 11.8 - PLOFAM (significant leaks): Complementary cumulative leak frequency distribution of
observed gas leaks at NCS and model prediction per equipment type for all installations in the NCS
of installation years. The total number of installation years for the NCS population dataset in the
period is 471

Figure 11.9 5 year floating average leak frequency per equipment year installations in the NCS
population dataset (62 installations being in operation in the period 2001-2014)
11.6 Randomness
An important aspect is the uncertainty with regard to prediction of large leaks. Large leaks in this
context are leaks having an initial leak rate 10 kg/s. The stochastic uncertainty can be assessed
based on the Poisson distribution. Application of the Poisson distribution for this purpose requires
that the leaks occur independently. This is not entirely true as it is reasonable to assume that
lesson learnt from incidents having some effect on the likelihood of leaks occurring in the future.
However, the effect of the dependency is considered to be small. Furthermore, this methodology
does not account for trend with time for leaks having an initial leak rate 0.1 kg/s in the period
(2001-2014). The methodology applies strictly for the given population for the given time period.
The method still provides useful information on the variance around the mean that is reasonable
to consider acceptable.
Figure 11.10 shows the Poisson distribution for the 15 observed large leaks in the NCS
population dataset over 802 installation years (i.e. 15 large leaks observed at 62 installations
located on the NCS and being in operation the period 01.01.2001-31.12.2014).
The green coloured bars approximate the 90 % confidence interval. This means that between 9
and 21 leaks can be expected with reasonable likelihood given a mean of 15 leaks. This means
that the occurrence of one or two large leaks in the near future does not lead to the conclusion
that the model ought to be disregarded.
Using the same approach for all leaks at NCS (mean of distribution equals 191 leaks having an
initial leak rate 0.1 kg/s) results in a confidence interval of 169 to 214 leaks. Hence, a significant
deviation from the mean could have been observed without rejecting the model estimate of the
total number of leaks at NCS installations.

Figure 11.10 Poisson distribution for 15 leaks having an initial leak rate equal to or larger than 10
kg/s in 802 installation years. The sum of the probability marked with green font is about 90 %.
The accumulated probability in each tail of the distribution is about 5 % (marked with blue font)
Figure 11.11 Poisson distribution for 191 leaks having an initial leak rate equal to or larger than
0.1 kg/s in 922 installation years. The sum of the probability marked with green font is about 90 %.
The accumulated probability in each tail of the distribution is about 5 % (marked with blue font)

12 Comparison PLOFAM and SHLFM
PLOFAM, is compared with the current frequency model commonly used in industry in Norway,
denoted SHLFM (Ref. /4/). The number of leaks predicted for the NCS population dataset (62
installations located on the NCS) using SHLFM is shown in Figure 12.3 through Figure 12.16. The
results presented are the total of the two main leak scenarios in the SHLFM model, the Full
pressure leak and Limited leak scenarios. Thus the contribution from the Zero pressure leak
scenario is disregarded.
The results demonstrate that:
The SHFLM model estimate of the total number of leaks in the NCS population dataset is
almost a factor of two higher than the observed number of leaks
The contribution from the various types of equipment predicted by the model is quite
different from the contribution observed in the data of leaks gathered from installations on
the NCS
The slope of the resulting SHLFM model complementary cumulative frequency distribution of
leaks with regard to initial leak rate is distinctly different than the observed historical
distribution. The deviation between the model prediction and the observed distribution
increases with increasing initial rate.
The frequency distributions for the NCS population dataset (62 installations located on the NCS)
obtained by use of SHLFM and PLOFAM is shown in Figure 12.1 and Figure 12.2. The results
display that the difference between the two models is considerable. The resulting quantitative fire
and explosion risk picture in a QRA for a typical installation on the NCS will be very different
based on PLOFAM opposed to SHLFM. The validation model has demonstrated that the model
denoted PLOFAM is able to predict the observed number of leaks at installations located on the
NCS in the period 2001-2014, whilst SHFLM will overpredict the observed number of leaks in the
same period. The difference is following from:
PLOFAM is based on historical data of leaks occurring at installations on the NCS in the
period 2001-2014. The last version of SHLFM was solely based on data of leaks occurring at
installations on the UKCS in the period 1992-2010. There has been a considerable decrease
in historical leak frequency at installations on the UKCS over this period. It has been shown
that the underlying hole size frequency distribution on the UKCS and NCS after year 2000 is
similar, and most likely the same statement is valid for the total leak frequency as well
Enhanced understanding of the quality of the data in the HCR database, which has provided
basis for implementing less conservatism in PLOFAM opposed to SHLFM to account for
uncertainty related to the data basis. This has been facilitated by that more data related to
the incidents in HCRD has been made available by HSE. Two items have been particularly
important in this regard. Firstly, the actual hole size for incidents where the hole size was
larger than 100 mm has been provided. Previously, it was stated that the hole size was > 100
mm in such cases. The additional information on large hole sizes has provided confidence in
estimation of a more accurate frequency for large leaks. Secondly, it has been found that the
population data (i.e. number of equipment years) in HCRD has not been updated after 2005,
which means that the estimated frequency for leaks extracted from HCRD will lead to an
excessive estimate of the leak frequency even for installations located on the UKCS
The mathematical formulation in PLOFAM enables an improved representation of the effect
of the equipment size on the hole size frequency distribution for the various equipment
types. In SHLFM, the capability in terms of capturing the shift in hole size distribution with
varying equipment size for a given equipment was less pronounced. In combination with
parameterisation of SHLFM outside the range of HCRD data for large holes led to estimation
of excessive frequency for large holes for all equipment sizes for all equipment type.
Moreover, as additional data on equipment size per incident in HCRD has been made
available to the project, it has been possible to develop and parameterize the equipment size
dependent model in PLOFAM.

Figure 12.1 Comparison SHLFM and PLOFAM. The frequency is the sum of frequency for gas and
liquid leaks. For the SHLFM model, only Full pressure leaks and Limited leaks are included. In
PLOFAM, only significant leaks are included
Figure 12.2 Comparison SHLFM and PLOFAM. The bars display ratio per leak category for the
total of gas and liquid leaks. For the SHLFM, only Full pressure leaks and Limited leaks are included.
In PLOFAM, only significant leaks are included

Figure 12.3 SHLFM: Observed number of leaks 0.1 kg/s versus model prediction for all

Figure 12.4 SHLFM: Observed number of leaks 0.1 kg/s versus model prediction per installation

Figure 12.5 SHLFM: Observed number of leaks 1 kg/s versus model prediction per installation for
the installations in the NCS population dataset (54 leaks observed at 62 installations being in

Figure 12.6 SHLFM: Observed number of leaks versus model prediction per leak category for all

Figure 12.7 - SHLFM: Observed number of leaks versus model prediction per equipment type for all
installations in the NCS population dataset (181 leaks observed at 62 installations being in opera-
tion in the period 01.01.2001-31.12.2014). The contribution from storage tanks and vents is not
modelled
Figure 12.8 SHLFM: Relative number of observed leaks and model prediction per equipment type
operation in the period 01.01.2001-31.12.2014). The coloured bars represent the individual in-

Figure 12.9 SHLFM: Complementary cumulative leak frequency distribution of observed leaks at
NCS and model prediction per equipment type for all installations in the NCS population dataset
total number of installation years for the NCS population dataset is 802. The contribution from
hoses is stemming from flexible piping
Figure 12.10 SHLFM: Normalized complementary cumulative leak frequency distribution of

aggregated normalized frequency equals 1.0). The contribution from hoses is stemming from
flexible piping

Figure 12.11 - SHLFM: Observed number of gas leaks versus model prediction per leak category for

Figure 12.12 - SHLFM: Observed number of liquid leaks versus model prediction per leak category
Figure 12.13 - SHLFM: Complementary cumulative distribution of leaks at NCS and model


13 MISOF benchmarking model and PLOFAM

The benchmark model for the installation denoted Installation A in the MISOF ignition model
(Ref. /5/) has been rerun to investigate the total effect on the fire and explosion probability of the
combined use of MISOF and PLOFAM. Installation A is a combined drilling and process plat-
form with a jacket substructure located on the Norwegian Continental Shelf that was set in
operation after year 2000.
Lloyd's Registers model for detailed explosion risk assessment (ExploRAM = Explosion Risk Assess-
ment Model) has been applied for calculation of the ignition probability resulting from exposure
to potential ignition sources inside the classified area the leak is originating from. Ignitions taking
place on the outside of the classified zone are quantified using a different exposure model purely
based on CFD calculations. In the latter case, the main concern is exposure to gas turbine air in-
takes. The ExploRAM model is in accordance with the best industry practice and NORSOK
standard, Ref. /6/. The first version of the model was presented in a paper at the ERA Technology
Conference "Safety on Offshore Installations" 30 November - 1 December 1999 (Ref. /7/).
The results based on the MISOF model combined with SHLFM are shown in Figure 13.1. The
results for the case with the MISOF model combined with PLOFAM are presented in Figure 13.2.
The results for the OLF ignition model (Ref. /8/) combined with the leak frequency models has not
been investigated. It is expected that the resulting explosion pressure frequency distribution for
the combined use of the OLF ignition model and the PLOFAM leak frequency model would be in
between the distributions displayed in Figure 13.1 and Figure 13.2.
The results demonstrate the effect of PLOFAM on the fire and explosion pressure is prominent.
fire and explosion frequency. For platforms like Installation A where special sources of ignition,
such as gas turbine air intakes, is a dominating factor, the effect of PLOFAM is more prominent
than for other installations. This is because the exposure probability for special sources of ignition
is dominated by large leaks. The relative reduction in leak frequency using PLOFAM instead of
SHLFM increases with increasing initial leak rate.

Figure 13.1 MISOF combined with SHLFM: Frequency distribution for explosion pressure for all

Figure 13.2 MISOF combined with PLOFAM: Frequency distribution for explosion pressure for all
14 Concluding remarks
A main overall conclusion obtained from running the parameterisation and validation process is
that the underlying hole size frequency distribution for equipment at installations located on the
NCS is similar to the distribution for equipment located on installations on the UKCS. The
differences may be explained by uncertainty related to both datasets (both the leaks and the
population data), limitations of the mathematical formulations and uncertainty associated with
the parameterization and validation methodology. Only a reduction of 20% (on average for all
parameters) of the frequency parameter ( ) was necessary to fit the observed frequency of
leaks at installations on the NCS (216 estimated with UKCS model versus 181 observed). The
major difference between the estimated leaks and the observed data stem from difference in
distribution in terms of type of equipment. Largely, the adjustment of the initial parameters
established based on data from installations on the UKCS required to obtain a model that is able
to describe the occurrence of leaks at installation on the NCS quite accurately can be considered
to be minor.
PLOFAM is able to:
Generate the total number of leaks at NCS in the period 01.01.2001 31.12.2014
Generate the split of the total number of leaks at NCS in the period 01.01.2001
31.12.2014 on significant and marginal leaks
Generate the number of significant gas leaks at NCS in the period 01.01.2007 31.12.2014
Generate the observed distribution of leaks with respect to initial leak rate for installations on
the NCS. The model predicts the individual contribution from significant and marginal leaks
as well as the frequency distribution for both leak scenarios

Generate the observed contribution to leaks originating from the different equipment types.
The model does also reproduce the observed distribution of leaks with respect to initial leak
rate for the equipment types most of the leaks at NCS are associated with (i.e. valves,
flanges, instruments and steel pipes).
The strategy has been to build a model that gives a best estimate for future leak frequencies, i.e.
to create an unbiased model without built in conservatism. It is observed a significant decreasing
trend in historical leak frequency with time for installations on the NCS in the period after year
2000. The period 2001-2014 is used as basis for the model, but the historical frequency for all
installations on NCS the last 5 year period is 40% less than the average for the period 2001-2014.
Hence, presuming that the number of leaks at the NCS in the future will follow the observed
frequency for the last 5 year period, PLOFAM is regarded to give robust results for future average
leak frequencies for installations on the NCS. The period 2001-2014 is used as basis for the
model to account for uncertainty in the data material and shifts in underlying casual factors (e.g.
emerging unknown degradation mechanisms due to age or changing operational conditions)
affecting the future trend in leaks occurring on installations on the NCS.
It should be mentioned that the historical leak frequency per installation at the NCS can vary
significantly from the NCS average, as a result of stochastic effects.
and how initial leak rates are modelled in the validation process. The validation model show that
The causes for the difference in model prediction with respect to fluid phase have not been
identified. An improved model prediction could have been obtained by defining separate model
parameters for equipment containing gas and liquid. However, no apparent explanation has been
found, and it has not been attempted to establish specific models for equipment containing gas
and equipment. It is recommended that future projects address this challenge.
PLOFAM has been compared with the commonly used leak frequency model denoted SHLFM
(Standardised Hydrocarbon Leak Frequencies Model, Ref. /4/). The difference between the leak
frequencies generated by the two models is considerable. These differences are explained by
effects following from properties of the new data material being available for development of
PLOFAM as well as new features of the mathematical framework enabling improved
representation of the equipment size for the various equipment types. The resulting quantitative
fire and explosion risk picture in a QRA for a typical installation on the NCS will be significantly
different based on PLOFAM opposed to SHLFM. The validation model has demonstrated that
PLOFAM is able to predict the observed number of leaks in the period 2001-2014 at installations
located on the NCS, whilst SHFLM will overpredict the observed number of leaks for the same
period.
The benchmark model in the MISOF (Ref. /5/) ignition model has been rerun to investigate the
total effect on the fire and explosion probability of the combined use of MISOF and PLOFAM. The
results demonstrate the effect of PLOFAM on the fire and explosion pressure is prominent. The
effect will depend on the type of ignition sources constituting the major contribution to the fire
and explosion frequency. For platforms where special sources of ignition, such as gas turbine air
intakes, are a dominating factor, the effect of PLOFAM relative to the use of SHLFM is more
prominent than for other modules. This is because the exposure probability for special sources of
ignition is dominated by large leaks. The relative reduction in leak frequency using PLOFAM
instead of SHLFM increases with increasing initial leak rate.

PLOFAM is validated towards data gathered from installations at NCS. However, it is
demonstrated that the PLOFAM hole size frequency distributions for equipment at installations
located on the NCS is similar to the distributions for equipment located on installations on the
UKCS. The model is therefore judged to be applicable also for installations on the UKCS.
Although the model is based on releases of hydrocarbons from process equipment on North Sea
offshore facilities, it is found reasonable to argue that the model is applicable to platforms and
land based facilities in other domains. This should be based on a specific assessment to qualify
use of the model in the particular domain. The important element to evaluate is whether the
properties of the equipment and operation conditions can be considered equivalent with what
are found generally on installations located in the North Sea.
15 Further work
A model that is expected to predict leaks at installations on the NCS has been developed. The
work in development of the model has identified some unresolved challenges that are suggested
to be addressed in future projects.
1) PLOFAM apply for estimation of leaks having an initial leak rate above 0.1 kg/s. In some
cases, e.g. leaks in small enclosures with poor ventilation rate, even smaller leak rates should
be assessed. The capability of the model to estimate the leak frequency for leaks with an
initial leak rate less than 0.1 kg/s should be investigated further.
2) Although the quality of the NCS population dataset is considered to be high, there are
aspects of the data that would increase the precision of the leak frequency model if enhan-
ced. Elements of particular interest in addition to general quality assurance and update of
equipment counts from QRAs are the effect of modifications implemented at installations,
number of hose operations, wells in operation, equipment counts of flanges and valves
associated with instrument connections and equipment counts of length of steel piping. The
population data basis for development of the model for leaks stemming from hoses used in
temporary operations is particularly uncertain. It is believed that the applied number of ave-
rage hose operations per installation used as basis for the validation of PLOFAM is somewhat
too low, which implies that the frequency for leaks from hoses most likely is conservative. It is
recommended that a future project upgrade the accuracy of the population dataset of hose
operations in order to improve the precision of the leak frequency originating from hoses
3) A complete understanding of the causes for the difference in model prediction of leaks in
terms of fluid phase has not been identified. An improved model prediction could have been
obtained by defining separate model parameters for equipment containing gas and liquid.
However, no apparent explanation has been found, and it has not been attempted to
establish specific models for equipment containing gas and equipment. It is recommended
that future projects address this challenge. One objective could be to improve the simplified
models used for estimation of initial leak rate for multi-phase leaks where estimation of the
correct density is challenging. Uncertainty when classifying leaks in terms of fluid phase is
also prominent, which calls for an improved methodology to be used under classification of
leaks in the accident investigation work
4) Many aspects of the model lack support in casual arguments. For instance, the trends in the
available data that describes the difference between various types of valves in terms of leak
frequency is not fully understood (e.g. the difference between actuated and manual valves).
It is recommended that future work seek to establish a better understanding of the corre-
lation between failure modes and the observed statistical data of leaks. FMEA (Failure Mode
and Effect Analysis) is suggested as tool for analysis of failure modes. Such studies would be
very useful for all equipment types to improve the fundament for the leak frequency model

5) The established mathematical framework is believed to not completely describe the behavi-
our of the underlying frequency hole size distributions. For the data available in this project,
the mathematical formulation has been proven to be adequate. Future studies should look
into possible improvements of the mathematical formulation. It is suggested that this is done
in parallel with studies of failure modes (see previous item 4)) that can enable an under-
standing of the fundamental properties of the underlying hole size probability density
function for the various equipment types
6) A general discrepancy between the estimated initial leak rate and hole size for leaks in HCRD
has not been understood. This should be investigated further in the future, which probably
also would increase understanding on how to improve modelling of initial leak rates in
general (see also item 3) above). More data that will shed light on this through the ongoing
upgrade project of the HCRD run by HSE in UK. The output from the upgrade project is also
expected to provide basis for enhanced understanding of the observed deviations between
data of leaks at installations on the NCS and the HCRD, both with regard to incidents (e.g.
difference in relative distribution between leaks stemming from various types of equipment)
and the population data (e.g. counts of equipment with few incidents (such as compressors,
atmospheric vessels), counts of instrument connections, counts of flanged surfaces versus
flanges co, of steel pipe etc.). A review of HCRD should be performed after the issue of the
updated version of HCRD
7) The fraction marginal leaks of leaks at the NCS are considered uncertain. It is judged that the
applied fraction in PLOFAM is somewhat conservative, i.e. somewhat too low. The fraction
should be revised when more data is available. The updated version of HCRD (see item 6)
above) will also improve the basis for setting the split between the significant and marginal
leak scenario
8) Considerable work has been put into establishing the leak frequency data for installations on
the NCS. It is recommended that the established data is maintained (both the database of
leaks and the corresponding population data) on a regular basis by a dedicated entity to
ensure effective studies in the future
9) It is recommended that the procedures for investigation of leaks is updated to ensure that
data parameters useful in modelling of leak frequencies is captured as part of the accident
investigation process (e.g. pressure, temperature, hole size, composition, inventory etc.). In
particular it would prove to very useful if the incidents are attributed to the correct equip-
ment type. A special area of interest is instrument connections. It would prove too valuable
to gather data that enable understanding of to what extent leaks from valves and flanges
related to instrument connections contributes to the overall distribution of leaks. This is a
factor of uncertainty in the data of leaks gathered from installations on the NCS as well as in
HCRD.
10) As described in TN-2 there is uncertainty related to what extent limited time in operation has
been taken into account in the equipment count database. This leads to a small uncertainty
in the established leak frequencies. In future updates, it is recommended that the equipment
count database is updated to ensure that no adjustments are made in the database to take
limited operational time into account.

16 References
th
/1/ Minutes of meeting, Workshop 4, September 15 2015, LR Consulting, Sandvika,
Norway.
/2/ Statoil, Guideline for application of the OMT model, Rev. 3.0, 2014-02-05.
/3/ Personal communication with Tor Eriksen, independent consultant specialist on compact
flanges.
1768, rev. 1, 16.01.2009.
/5/ Scandpower AS: Modelling of ignition sources on offshore oil and gas facilities, Report
th
No. 102657/R1, September 18 2014.
/6/ The NORSOK Standard for Risk and Emergency Preparedness Analyses Z-013, Rev. 2,
September 2001.
/7/ Safety on Offshore Installations - Conference Proceedings, 30 November - 1 December

1999, ERA Technology.
/8/ Scandpower AS: "Ignition modelling in risk analysis", report no. 89.390.008/R1, March
2007.

Appendix A
Model Parameters

1 Introduction .................................................................................................................................. A1

1 Introduction
The parameters for the equipment types for PLOFAM are given in addition to the parameters for
the models denoted UKCS model and NCS model in this Appendix.
The parameters for compressor and pump (only provided for PLOFAM) were developed as part of
the validation model since the type of compressor and pump (i.e. reciprocal vs. centrifugal) were
unknown for some equipment counts in the NCS population dataset. As such, the amalgamated
model for pump and compressor is based on the relative distribution in number of objects being
centrifugal and reciprocating (for pumps and compressors individually) on platforms in the North
Sea.
The PLOFAM parameters for the specific valve types manual valve, actuated valve and ESD valve is
only to be used in special studies where the type of valve is known for all valves of the system
being studied. In general it is recommended to use the amalgamated model for valve.

Table 1.1 - UKCS model parameters (- indicate zero)
Equipment type Scenario A_0 m_0 A_D m_D B_D F_hist

Air-cooled heat exchanger Significant leak 1.00 - - - 2.70E-02 5.67E-04
Air-cooled heat exchanger Marginal leak - - - - - -
Atmospheric vessel Significant leak 1.00 - - - 1.50E-01 8.60E-03
Atmospheric vessel Marginal leak 1.00 - - - 1.50E-01 3.00E-03
Centrifugal compressor Significant leak 1.00 - - - 1.20E-02 3.30E-03
Centrifugal compressor Marginal leak 1.00 - - - 1.20E-02 5.80E-03
Centrifugal pump Significant leak 1.00 - - - 3.10E-04 1.05E-02
Centrifugal pump Marginal leak 1.00 - - - 3.10E-04 3.70E-03
Filter Significant leak 1.00 - - - 1.00E-02 2.20E-03
Filter Marginal leak 1.00 - - - 1.00E-02 7.00E-04
Flexible pipe Significant leak 1.00 - - - 7.80E-02 1.50E-04
Flexible pipe Marginal leak 1.00 - - - 7.80E-02 1.00E-04
Gas lift well Significant leak 1.00 - - - 1.70E-02 1.70E-03
Gas lift well Marginal leak 1.00 - - - 1.70E-02 1.70E-03
Instrument Significant leak 0.30 0.40 3.45 -1.25 8.60E-04 1.80E-04
Instrument Marginal leak 0.30 0.40 3.45 -1.25 8.60E-04 1.00E-04
Pig trap Significant leak 1.00 - - - 1.20E-02 4.40E-03
Pig trap Marginal leak 1.00 - - - 1.20E-02 2.20E-03
Plate heat exchanger Significant leak 1.00 - - - 5.00E-03 5.40E-03
Plate heat exchanger Marginal leak 1.00 - - - 5.00E-03 1.60E-03
Process vessel Significant leak 1.00 - - - 5.40E-02 6.90E-04
Process vessel Marginal leak 1.00 - - - 5.40E-02 1.03E-03
Producing well Significant leak 1.00 - - - 3.80E-02 7.40E-05

Producing well Marginal leak 1.00 - - - 3.80E-02 4.40E-04
Reciprocating compressor Significant leak 1.00 - - - 4.20E-02 2.30E-02
Reciprocating compressor Marginal leak 1.00 - - - 4.20E-02 2.10E-02
Reciprocating pump Significant leak 1.00 - - - 5.60E-02 5.60E-03
Reciprocating pump Marginal leak 1.00 - - - 5.60E-02 1.90E-03
Shell and tube heat exchanger Significant leak 1.00 - - - 2.70E-02 1.40E-03
Shell and tube heat exchanger Marginal leak 1.00 - - - 2.70E-02 5.00E-04
Standard flange Significant leak 0.17 0.40 2.00 -1.00 1.50E-02 1.70E-05
Standard flange Marginal leak 0.17 0.40 2.00 -1.00 1.50E-02 1.10E-05
Steel pipe Significant leak 4.20 -0.30 17.60 -1.75 2.30E-03 4.40E-05
Steel pipe Marginal leak 4.20 -0.30 17.60 -1.75 2.30E-03 1.90E-05
Valve Significant leak 0.19 0.40 2.65 -1.25 6.60E-04 7.20E-05
Valve Marginal leak 0.19 0.40 2.65 -1.25 6.60E-04 4.10E-05

Table 1.2 - NCS model parameters (- indicate zero)

Air-cooled heat exchanger Significant leak - - - - - -
Atmospheric vessel Marginal leak 1.00 - - - 1.50E-01 -
Centrifugal compressor Marginal leak 1.00 - - - 1.20E-02 -
Centrifugal pump Marginal leak 1.00 - - - 3.10E-04 -
Compact flange Significant leak 1.00 - - - 1.00E-03 3.30E-06
Compact flange Marginal leak 1.00 - - - 1.00E-03 -
Filter Marginal leak 1.00 - - - 1.00E-02 -
Flexible pipe Marginal leak 1.00 - - - 7.80E-02 -
Hose Significant leak 1.00 - - - 2.00E-01 1.00E-04
Hose Marginal leak 1.00 - - - 2.00E-01 2.00E-05
Pig trap Marginal leak 1.00 - - - 1.20E-02 -

Plate heat exchanger Marginal leak 1.00 - - - 5.00E-03 -
Process vessel Marginal leak 1.00 - - - 5.40E-02 -
Reciprocating compressor Marginal leak 1.00 - - - 4.20E-02 -
Reciprocating pump Marginal leak 1.00 - - - 5.60E-02 -
Shell and tube heat exchanger Marginal leak 1.00 - - - 2.70E-02 -

Table 1.3 - PLOFAM parameters (- indicate zero)

Air-cooled heat exchanger Significant leak 1.00 - - - 2.70E-02 5.67E-04
Atmospheric vessel Marginal leak 1.00 - - - 1.50E-01 -
Centrifugal compressor Marginal leak 1.00 - - - 4.50E-03 -
Centrifugal pump Marginal leak 1.00 - - - 5.00E-07 -
Compact flange Significant leak 1.00 - - - 1.00E-03 3.00E-06
Compact flange Marginal leak 1.00 - - - 1.00E-03 -
Filter Marginal leak 1.00 - - - 1.50E-04 -
Flexible pipe Marginal leak 1.00 - - - 7.80E-02 -
Hose Significant leak 1.00 - - - 2.00E-01 1.00E-04
Hose Marginal leak 1.00 - - - 2.00E-01 2.00E-05
Pig trap Marginal leak 1.00 - - - 1.20E-02 -

Plate heat exchanger Marginal leak 1.00 - - - 5.00E-04 -
Process vessel Marginal leak 1.00 - - - 9.00E-04 -
Reciprocating compressor Marginal leak 1.00 - - - 6.00E-03 -
Reciprocating pump Marginal leak 1.00 - - - 1.00E-02 -
Shell and tube heat exchanger Marginal leak 1.00 - - - 4.30E-03 -

Table 1.4 - PLOFAM parameters for additional equipment types. The parameters for compressor and pump where developed as part of the validation model as
the type of compressor and pump (i.e. reciprocal vs. centrifugal) where unknown for some equipment counts in the NCS population dataset. As such, the
amalgamated model for pump and compressor is based on the relative distribution in number of objects being centrifugal and reciprocating in the population
data. (- indicate zero)

Compressor Significant leak 1.00 - - - 4.60E-03 3.00E-03
Compressor Marginal leak 1.00 - - - 4.60E-03 -
Pump Significant leak 1.00 - - - 1.00E-06 8.40E-03
Pump Marginal leak 1.00 - - - 1.00E-06 -
ESD valve Significant leak 1.00 - 2.90 -1.75 7.25E-04 6.00E-04
ESD valve Marginal leak 1.00 - 2.90 -1.75 7.25E-04 7.00E-05
Manual valve Significant leak 0.13 0.50 2.40 -1.25 1.50E-02 1.50E-04
Manual valve Marginal leak 0.13 0.50 2.40 -1.25 1.50E-02 1.70E-05
Actuated valve Significant leak 1.00 - 2.90 -1.75 7.25E-04 1.00E-03
Actuated valve Marginal leak 1.00 - 2.90 -1.75 7.25E-04 1.10E-04

Appendix B
UKCS model: data basis and

intermediate results

1 Introduction ...................................................................................................................................B1
2 Complementary cumulative equipment size distributions ...............................................................B2
3 Complementary cumulative frequency hole size distributions.........................................................B5
4 Full bore hole fraction ..................................................................................................................B31

1 Introduction
This Appendix contains:
Complementary cumulative equipment size distributions based on the NCS population
dataset (see TN-2)
Fitted complementary cumulative hole size frequency distribution for all equipment types
based on corresponding distribution extracted from HCR (see TN-3). The fitting is performed
in a spreadsheet where all required data is compiled. The fitted distributions can be reprodu-
ced from the UKCS model parameters presented in Appendix A (to this TN). For some
equipment types, also the result using a different filter than the filter used to estimate
parameters the parameters in the UKCS model are shown
Full bore rupture fraction per equipment dimension for valves, standard flanges and steel
pipes (extracted from HCRD)

2 Complementary cumulative equipment size distributions
Table 2.1 - Complementary cumulative equipment size distribution per equipment type. The data is compiled from on the NCS population dataset
described in TN-2
Lower boundary Atmospheric Centrifugal Centrifugal

Actuated valve Compact flange Compressor Filter Instrument Manual valve Pig trap
of class (inch) vessel compressor pump
0 99.99 % 100.00 % 100.00 % 100.00 % 100.00 % 100.00 % 100.00 % 100.00 % 100.00 % 100.00 %
0.375 98.57 % 100.00 % 100.00 % 100.00 % 100.00 % 100.00 % 100.00 % 99.93 % 99.89 % 100.00 %
0.625 98.32 % 100.00 % 100.00 % 100.00 % 100.00 % 100.00 % 100.00 % 90.57 % 90.71 % 100.00 %
0.875 87.91 % 100.00 % 100.00 % 100.00 % 99.67 % 100.00 % 98.91 % 20.01 % 87.26 % 100.00 %
1.25 87.65 % 66.67 % 100.00 % 98.09 % 99.67 % 100.00 % 97.27 % 2.04 % 64.76 % 99.95 %
1.75 67.13 % 66.67 % 100.00 % 97.77 % 99.67 % 99.66 % 97.27 % 1.79 % 64.18 % 99.95 %
2.5 56.75 % 0.00 % 100.00 % 88.24 % 98.50 % 98.99 % 93.64 % 0.34 % 37.66 % 99.91 %
3.5 46.98 % 0.00 % 99.60 % 81.81 % 98.50 % 98.65 % 89.82 % 0.17 % 30.09 % 99.91 %
4.5 46.47 % 0.00 % 97.21 % 74.76 % 73.54 % 96.62 % 86.00 % 0.10 % 23.96 % 99.91 %
5.5 34.29 % 0.00 % 97.21 % 74.76 % 73.54 % 96.62 % 86.00 % 0.10 % 23.88 % 99.91 %
6.5 34.29 % 0.00 % 92.82 % 64.72 % 60.25 % 90.19 % 64.72 % 0.10 % 16.01 % 95.82 %
7.5 22.23 % 0.00 % 92.82 % 64.72 % 60.25 % 90.19 % 64.72 % 0.10 % 16.01 % 95.82 %
8.5 22.22 % 0.00 % 78.85 % 55.48 % 42.55 % 77.33 % 60.36 % 0.10 % 9.82 % 93.09 %
9.5 15.24 % 0.00 % 78.85 % 55.48 % 42.55 % 77.33 % 60.36 % 0.10 % 9.80 % 93.09 %
10.5 9.87 % 0.00 % 65.08 % 44.99 % 14.20 % 64.31 % 58.72 % 0.10 % 6.99 % 79.26 %
11.5 9.87 % 0.00 % 65.08 % 44.99 % 14.20 % 64.31 % 58.72 % 0.10 % 6.99 % 79.26 %
13.5 8.00 % 0.00 % 49.32 % 37.68 % 7.00 % 49.59 % 54.90 % 0.03 % 4.65 % 73.80 %
14.5 8.00 % 0.00 % 35.12 % 32.91 % 4.90 % 37.55 % 52.44 % 0.03 % 3.43 % 72.39 %
15.5 4.33 % 0.00 % 35.12 % 32.91 % 4.90 % 37.55 % 52.44 % 0.03 % 3.42 % 72.39 %
16.5 4.33 % 0.00 % 15.16 % 11.29 % 3.03 % 20.64 % 10.43 % 0.03 % 1.64 % 25.92 %
17.5 3.92 % 0.00 % 15.16 % 11.29 % 3.03 % 20.64 % 10.43 % 0.03 % 1.64 % 25.92 %
19.5 1.34 % 0.00 % 13.17 % 10.02 % 2.10 % 11.50 % 9.89 % 0.03 % 1.45 % 20.46 %
21 1.34 % 0.00 % 5.19 % 7.15 % 2.10 % 4.74 % 7.71 % 0.00 % 0.40 % 15.00 %
22.5 1.33 % 0.00 % 5.19 % 7.15 % 2.10 % 4.74 % 7.71 % 0.00 % 0.40 % 12.28 %
23.5 0.70 % 0.00 % 5.19 % 7.15 % 2.10 % 4.74 % 7.71 % 0.00 % 0.40 % 12.28 %
25 0.70 % 0.00 % 2.79 % 2.19 % 0.00 % 2.37 % 5.18 % 0.00 % 0.09 % 12.28 %
27 0.61 % 0.00 % 2.79 % 2.19 % 0.00 % 2.37 % 5.18 % 0.00 % 0.09 % 12.28 %
29 0.21 % 0.00 % 2.39 % 2.19 % 0.00 % 2.03 % 5.18 % 0.00 % 0.09 % 10.91 %
31 0.17 % 0.00 % 0.00 % 0.00 % 0.00 % 0.00 % 0.00 % 0.00 % 0.00 % 9.55 %
34 0.08 % 0.00 % 0.00 % 0.00 % 0.00 % 0.00 % 0.00 % 0.00 % 0.00 % 9.55 %
38 0.00 % 0.00 % 0.00 % 0.00 % 0.00 % 0.00 % 0.00 % 0.00 % 0.00 % 6.82 %
40 0.00 % 0.00 % 0.00 % 0.00 % 0.00 % 0.00 % 0.00 % 0.00 % 0.00 % 0.00 %

described in TN-2. L indicates HCR equipment size category >11, M indicates HCR equipment size category 3-11 and S HCR equipment size
category 3.
Lower boundary Reciprocating Reciprocating Shell and tube Shell side heat Tube side heat
Process vessel Pump Standard flange Valve Valve L Valve M Valve S
of class (inch) compressor pump heat exchanger exchanger exchanger
0 100.00 % 100.00 % 100.00 % 100.00 % 100.00 % 100.00 % 100.00 % 100.00 % 100.00 % 100.00 % 100.00 % 100.00 %
0.375 100.00 % 100.00 % 100.00 % 100.00 % 100.00 % 100.00 % 99.98 % 100.00 % 99.96 % 100.00 % 100.00 % 100.00 %
0.625 100.00 % 99.43 % 100.00 % 87.57 % 100.00 % 100.00 % 95.82 % 100.00 % 92.88 % 100.00 % 100.00 % 89.36 %
0.875 99.88 % 99.43 % 100.00 % 87.57 % 100.00 % 100.00 % 93.06 % 100.00 % 87.23 % 100.00 % 100.00 % 75.99 %
1.25 99.19 % 97.72 % 100.00 % 87.57 % 100.00 % 100.00 % 81.97 % 99.94 % 72.64 % 100.00 % 100.00 % 60.67 %
1.75 99.19 % 97.44 % 93.75 % 87.57 % 100.00 % 100.00 % 80.92 % 99.94 % 71.57 % 100.00 % 100.00 % 58.14 %
2.5 97.51 % 88.90 % 81.25 % 87.30 % 100.00 % 97.88 % 56.58 % 99.19 % 45.21 % 100.00 % 100.00 % 14.49 %
3.5 94.97 % 83.15 % 81.25 % 87.30 % 100.00 % 96.83 % 47.47 % 97.31 % 36.54 % 100.00 % 100.00 % 0.00 %
4.5 90.91 % 75.70 % 81.25 % 62.45 % 100.00 % 94.71 % 39.91 % 94.69 % 29.79 % 100.00 % 72.62 % 0.00 %
5.5 90.91 % 75.70 % 81.25 % 62.45 % 100.00 % 94.71 % 39.59 % 94.69 % 29.49 % 100.00 % 70.91 % 0.00 %
6.5 82.31 % 66.72 % 37.50 % 62.45 % 97.78 % 82.54 % 29.06 % 84.57 % 20.58 % 100.00 % 33.54 % 0.00 %
7.5 82.31 % 66.72 % 37.50 % 62.45 % 97.78 % 82.54 % 29.06 % 84.57 % 20.58 % 100.00 % 33.54 % 0.00 %
8.5 74.63 % 57.02 % 25.00 % 62.14 % 88.89 % 73.54 % 19.41 % 77.81 % 12.81 % 100.00 % 0.16 % 0.00 %
9.5 74.63 % 57.02 % 25.00 % 62.14 % 88.89 % 73.54 % 19.39 % 77.81 % 12.78 % 100.00 % 0.00 % 0.00 %
10.5 66.48 % 45.65 % 18.75 % 37.28 % 82.22 % 59.26 % 13.22 % 68.25 % 8.33 % 60.55 % 0.00 % 0.00 %
11.5 66.48 % 45.65 % 18.75 % 37.28 % 82.22 % 59.26 % 13.22 % 68.25 % 8.33 % 60.55 % 0.00 % 0.00 %
13.5 60.49 % 39.11 % 6.25 % 37.28 % 82.22 % 40.74 % 9.09 % 59.63 % 5.39 % 38.40 % 0.00 % 0.00 %
14.5 54.34 % 34.55 % 6.25 % 37.28 % 71.11 % 34.92 % 7.09 % 52.51 % 4.08 % 28.69 % 0.00 % 0.00 %
15.5 54.34 % 34.55 % 6.25 % 37.28 % 71.11 % 34.92 % 7.09 % 52.51 % 4.07 % 28.69 % 0.00 % 0.00 %
16.5 30.05 % 14.65 % 6.25 % 37.28 % 71.11 % 22.75 % 4.21 % 16.92 % 2.08 % 14.91 % 0.00 % 0.00 %
17.5 30.05 % 14.65 % 6.25 % 37.28 % 71.11 % 22.75 % 4.20 % 16.92 % 2.07 % 14.75 % 0.00 % 0.00 %
19.5 22.19 % 10.67 % 6.25 % 37.28 % 8.89 % 15.87 % 3.06 % 13.17 % 1.53 % 8.40 % 0.00 % 0.00 %
21 15.86 % 8.11 % 6.25 % 37.28 % 0.00 % 10.58 % 1.06 % 3.80 % 0.52 % 3.33 % 0.00 % 0.00 %
22.5 15.68 % 8.11 % 6.25 % 37.28 % 0.00 % 10.58 % 1.05 % 3.80 % 0.51 % 3.23 % 0.00 % 0.00 %
23.5 15.68 % 8.11 % 6.25 % 37.28 % 0.00 % 10.58 % 1.05 % 3.80 % 0.51 % 3.23 % 0.00 % 0.00 %
25 4.72 % 1.96 % 0.00 % 0.00 % 0.00 % 0.00 % 0.28 % 1.17 % 0.18 % 1.04 % 0.00 % 0.00 %
27 4.60 % 1.96 % 0.00 % 0.00 % 0.00 % 0.00 % 0.28 % 1.17 % 0.18 % 1.04 % 0.00 % 0.00 %
29 3.79 % 1.96 % 0.00 % 0.00 % 0.00 % 0.00 % 0.24 % 1.17 % 0.16 % 0.97 % 0.00 % 0.00 %
31 0.84 % 0.00 % 0.00 % 0.00 % 0.00 % 0.00 % 0.06 % 0.00 % 0.04 % 0.23 % 0.00 % 0.00 %
34 0.49 % 0.00 % 0.00 % 0.00 % 0.00 % 0.00 % 0.05 % 0.00 % 0.03 % 0.23 % 0.00 % 0.00 %
38 0.20 % 0.00 % 0.00 % 0.00 % 0.00 % 0.00 % 0.01 % 0.00 % 0.01 % 0.00 % 0.00 % 0.00 %
40 0.00 % 0.00 % 0.00 % 0.00 % 0.00 % 0.00 % 0.00 % 0.00 % 0.00 % 0.00 % 0.00 % 0.00 %

described in TN-2. L indicates HCR equipment size category >11, M indicates HCR equipment size category 3-11 and S HCR equipment size
category 3.
Lower boundary Actuated valve Actuated valve Actuated valve Standard flange Standard flange Standard flange
Manual valve L Manual valve M Manual valve S Steel pipe
of class (inch) L M S L M S
0 100.00 % 100.00 % 100.00 % 100.00 % 100.00 % 100.00 % 100.00 % 100.00 % 100.00 % 100.00 %
0.375 100.00 % 100.00 % 99.98 % 100.00 % 100.00 % 99.85 % 100.00 % 100.00 % 99.96 % 99.00 %
0.625 100.00 % 100.00 % 96.68 % 100.00 % 100.00 % 86.71 % 100.00 % 100.00 % 92.03 % 98.00 %
0.875 100.00 % 100.00 % 96.11 % 100.00 % 100.00 % 81.77 % 100.00 % 100.00 % 86.78 % 96.00 %
1.25 100.00 % 100.00 % 72.04 % 100.00 % 100.00 % 49.60 % 100.00 % 100.00 % 65.68 % 95.00 %
1.75 100.00 % 100.00 % 71.44 % 100.00 % 100.00 % 48.76 % 100.00 % 100.00 % 63.68 % 95.00 %
2.5 100.00 % 100.00 % 24.01 % 100.00 % 100.00 % 10.82 % 100.00 % 100.00 % 17.35 % 83.00 %
3.5 100.00 % 100.00 % 0.00 % 100.00 % 100.00 % 0.00 % 100.00 % 100.00 % 0.00 % 74.00 %
4.5 100.00 % 71.70 % 0.00 % 100.00 % 69.79 % 0.00 % 100.00 % 73.07 % 0.00 % 74.00 %
5.5 100.00 % 70.25 % 0.00 % 100.00 % 69.39 % 0.00 % 100.00 % 71.94 % 0.00 % 65.00 %
6.5 100.00 % 34.97 % 0.00 % 100.00 % 30.59 % 0.00 % 100.00 % 34.44 % 0.00 % 65.00 %
7.5 100.00 % 34.97 % 0.00 % 100.00 % 30.59 % 0.00 % 100.00 % 34.44 % 0.00 % 54.00 %
8.5 100.00 % 0.03 % 0.00 % 100.00 % 0.13 % 0.00 % 100.00 % 0.08 % 0.00 % 54.00 %
9.5 100.00 % 0.00 % 0.00 % 100.00 % 0.00 % 0.00 % 100.00 % 0.00 % 0.00 % 43.00 %
10.5 68.59 % 0.00 % 0.00 % 71.37 % 0.00 % 0.00 % 68.17 % 0.00 % 0.00 % 43.00 %
11.5 68.59 % 0.00 % 0.00 % 71.37 % 0.00 % 0.00 % 68.17 % 0.00 % 0.00 % 32.00 %
13.5 44.44 % 0.00 % 0.00 % 47.49 % 0.00 % 0.00 % 46.87 % 0.00 % 0.00 % 24.00 %
14.5 35.99 % 0.00 % 0.00 % 35.04 % 0.00 % 0.00 % 36.58 % 0.00 % 0.00 % 24.00 %
15.5 35.99 % 0.00 % 0.00 % 34.93 % 0.00 % 0.00 % 36.58 % 0.00 % 0.00 % 21.00 %
16.5 19.49 % 0.00 % 0.00 % 16.78 % 0.00 % 0.00 % 21.71 % 0.00 % 0.00 % 21.00 %
17.5 19.49 % 0.00 % 0.00 % 16.78 % 0.00 % 0.00 % 21.68 % 0.00 % 0.00 % 21.00 %
19.5 17.63 % 0.00 % 0.00 % 14.80 % 0.00 % 0.00 % 15.80 % 0.00 % 0.00 % 15.20 %
21 6.02 % 0.00 % 0.00 % 4.08 % 0.00 % 0.00 % 5.47 % 0.00 % 0.00 % 1.00 %
22.5 6.01 % 0.00 % 0.00 % 4.07 % 0.00 % 0.00 % 5.40 % 0.00 % 0.00 % 0.10 %
23.5 5.96 % 0.00 % 0.00 % 4.04 % 0.00 % 0.00 % 5.39 % 0.00 % 0.00 % 0.10 %
25 3.16 % 0.00 % 0.00 % 0.93 % 0.00 % 0.00 % 1.45 % 0.00 % 0.00 % 0.00 %
27 3.16 % 0.00 % 0.00 % 0.93 % 0.00 % 0.00 % 1.45 % 0.00 % 0.00 % 0.00 %
29 2.74 % 0.00 % 0.00 % 0.87 % 0.00 % 0.00 % 1.22 % 0.00 % 0.00 % 0.00 %
31 0.93 % 0.00 % 0.00 % 0.03 % 0.00 % 0.00 % 0.30 % 0.00 % 0.00 % 0.00 %
34 0.74 % 0.00 % 0.00 % 0.03 % 0.00 % 0.00 % 0.27 % 0.00 % 0.00 % 0.00 %
38 0.37 % 0.00 % 0.00 % 0.03 % 0.00 % 0.00 % 0.06 % 0.00 % 0.00 % 0.00 %
40 0.00 % 0.00 % 0.00 % 0.00 % 0.00 % 0.00 % 0.00 % 0.00 % 0.00 % 0.00 %

3 Complementary cumulative frequency hole size distributions
Figure 3.1 - Atmospheric vessel: resulting fit to the complementary cumulative frequency hole size distribution extracted from the HCRD (hole sizes
based on filter 1a, frequency based on period 2001 Q1 2015 TN-3). The vertical axis represents the weighted F(d) for all equipment dimensions

Figure 3.2 - Centrifugal compressor: resulting fit to the complementary cumulative frequency hole size distribution extracted from the HCRD (hole sizes
based on filter 1b, frequency based on period 2001 Q1 2015 TN-3). The vertical axis represents the weighted F(d) for all equipment dimensions

Figure 3.3 - Centrifugal pump: resulting fit to the complementary cumulative frequency hole size distribution extracted from the HCRD (hole sizes

Figure 3.4 - Compressors (total of centrifugal and reciprocating compressors): resulting fit to the complementary cumulative frequency hole size
distribution extracted from the HCRD (hole sizes based on filter 1a, frequency based on period 1992 Q1 2015 TN-3). The vertical axis represents the
weighted F(d) for all equipment dimensions

Figure 3.5 - Compressors (total of centrifugal and reciprocating compressors): resulting fit to the complementary cumulative frequency hole size
distribution extracted from the HCRD (hole sizes based on filter 1b, frequency based on period 2001 Q1 2015 TN-3). The vertical axis represents the

Figure 3.6 - Filter: resulting fit to the complementary cumulative frequency hole size distribution extracted from the HCRD (hole sizes based on filter 1a,
frequency based on period 1992 Q1 2015 TN-3). The vertical axis represents the weighted F(d) for all equipment dimensions

Figure 3.7 - Filter: resulting fit to the complementary cumulative frequency hole size distribution extracted from the HCRD (hole sizes based on filter 1b,
frequency based on period 2001 Q1 2015 TN-3). The vertical axis represents the weighted F(d) for all equipment dimensions

Figure 3.8 - Flexible pipe: resulting fit to the complementary cumulative frequency hole size distribution extracted from the HCRD (hole sizes based on
filter 1a, frequency based on period 1992 Q1 2015 TN-3). The vertical axis represents the weighted F(d) for all equipment dimensions

Figure 3.9 - Gas lift well (equipment dimension per definition set to 6 inch): resulting fit to the complementary cumulative frequency hole size

Figure 3.10 - Instrument: resulting fit to the complementary cumulative frequency hole size distribution extracted from the HCRD (hole sizes based on

Figure 3.11 - Instrument: resulting fit to the complementary cumulative frequency hole size distribution extracted from the HCRD (hole sizes based on
filter 1b, frequency based on period 2001 Q1 2015 TN-3). The vertical axis represents the weighted F(d) for all equipment dimensions

Figure 3.12 - Pig trap: resulting fit to the complementary cumulative frequency hole size distribution extracted from the HCRD (hole sizes based on

Figure 3.13 - Plate heat exchanger: resulting fit to the complementary cumulative frequency hole size distribution extracted from the HCRD (hole sizes

Figure 3.14 - Process vessel: resulting fit to the complementary cumulative frequency hole size distribution extracted from the HCRD (hole sizes based
on filter 1a, frequency based on period 2001 Q1 2015 TN-3). The vertical axis represents the weighted F(d) for all equipment dimensions

Figure 3.15 - Producing well (equipment dimension per definition set to 12 inch): resulting fit to the complementary cumulative frequency hole size

Figure 3.16 - Pump (total for centrifugal and reciprocal pumps): resulting fit to the complementary cumulative frequency hole size distribution
extracted from the HCRD (hole sizes based on filter 1b, frequency based on period 2001 Q1 2015 TN-3). The vertical axis represents the weighted F(d)
for all equipment dimensions

Figure 3.17 - Reciprocating compressor: resulting fit to the complementary cumulative frequency hole size distribution extracted from the HCRD (hole
sizes based on filter 1b, frequency based on period 2001 Q1 2015 TN-3). The vertical axis represents the weighted F(d) for all equipment dimensions

Figure 3.18 - Reciprocating pump: resulting fit to the complementary cumulative frequency hole size distribution extracted from the HCRD (hole sizes

Figure 3.19 - Shell and tube heat exchanger (total of Shell side and Tube side heat exchanger): resulting fit to the complementary cumulative
frequency hole size distribution extracted from the HCRD (hole sizes based on filter 1b, frequency based on period 2001 Q1 2015 TN-3). The vertical
axis represents the weighted F(d) for all equipment dimensions

Figure 3.20 - Shell side heat exchanger: resulting fit to the complementary cumulative frequency hole size distribution extracted from the HCRD (hole

Figure 3.21 - Standard flange: resulting fit to the complementary cumulative frequency hole size distribution extracted from the HCRD (hole sizes

Figure 3.22 - Standard flange with alpha = 0.9 (see TN-5): resulting fit to the complementary cumulative frequency hole size distribution extracted
from the HCRD (hole sizes based on filter 1b, frequency based on period 2001 Q1 2015 TN-3). The vertical axis represents the weighted F(d) for all
equipment dimensions

Figure 3.23 - Tube side heat exchanger: resulting fit to the complementary cumulative frequency hole size distribution extracted from the HCRD (hole

Figure 3.24 - Steel pipe: resulting fit to the complementary cumulative frequency hole size distribution extracted from the HCRD (hole sizes based on
filter 1b, frequency based on period 2001 Q1 2015 TN-3). The vertical axis represents the weighted F(d) for all equipment dimensions

Figure 3.25 - Manual valve: resulting fit to the complementary cumulative frequency hole size distribution extracted from the HCRD (hole sizes based
on filter 1b, frequency based on period 2001 Q1 2015 TN-3). The vertical axis represents the weighted F(d) for all equipment dimensions

Figure 3.26 - Actuated valve: resulting fit to the complementary cumulative frequency hole size distribution extracted from the HCRD (hole sizes based
on filter 1b, frequency based on period 2001 Q1 2015 TN-3). The vertical axis represents the weighted F(d) for all equipment dimensions

4 Full bore hole fraction
Table 4.1 - Valves: Full bore hole fraction extracted from HCRD
Full bore hole Full bore hole

Equipment fraction, floating Equipment fraction, floating
fraction, Valve, HCR fraction, Valve, HCR
dimension (inch) average HCR, Valve dimension (inch) average HCR, Valve
2001-2015 1992-Q12015
01-15 92-15
0.1 1.0 0.1 1.0
0.2 - 0.5 0.2 - 0.4

0.3 0.4 0.1 0.3 0.3 0.1
0.4 - 0.1 0.4 - 0.1
0.4 - 0.1 0.4 - 0.1
0.5 0.3 0.2 0.5 0.3 0.3
0.5 0.4 0.2 0.5 0.5 0.3
0.7 - 0.1 0.7 - 0.2
0.7 - 0.1 0.7 - 0.1
0.8 0.2 0.2 0.8 0.4 0.5
1.0 0.3 0.2 0.8 1.0 0.6

1.0 0.1 0.1 1.0 0.3 0.5
1.4 - 0.0 1.0 0.1 0.1
1.5 - - 1.4 - 0.1
2.0 - 0.0 1.5 0.1 0.0
2.0 0.1 0.0 2.0 - 0.1
2.0 - 0.0 2.0 0.1 0.0
3.0 - - 2.0 - 0.0
4.0 - - 3.0 0.0 0.0

5.1 - - 3.9 - 0.0
5.1 - 0.0 4.0 0.0 0.0
6.0 0.0 0.0 5.1 - 0.0
6.0 - 0.0 5.1 - -
8.0 - - 5.9 - 0.0
10.0 - - 6.0 0.1 0.0
12.0 - - 6.0 - 0.0
14.0 - - 8.0 0.1 0.1

16.0 - - 10.0 0.1 0.1
18.0 - - 12.0 - 0.0
20.0 - - 14.0 - -
30.0 - - 15.7 - -
32.0 - - 16.0 - -
36.0 - 18.0 - -
20.0 - -
30.0 - -
32.0 - -
36.0 -

Table 4.2 - Standard flange: Full bore hole fraction extracted from HCRD

fraction, floating fraction, floating
Equipment fraction, Standard Equipment fraction, Standard
average HCR, average HCR,
dimension (inch) flange, HCR 2001- dimension (inch) flange, HCR 1992-
Standard flange 01- Standard flange 92-
2015 Q12015
15 15
0.2 - 0.2 -
0.3 - 0.2 0.3 - 0.1

0.5 0.5 0.2 0.5 0.2 0.1
0.5 - 0.2 0.5 - 0.1
0.7 - - 0.7 - 0.1
0.8 - - 0.8 0.2 0.1
1.0 - 0.0 1.0 - 0.2
1.0 0.1 0.0 1.0 0.3 0.2
1.5 - 0.0 1.5 0.1 0.2
2.0 - - 2.0 0.0 0.1
2.0 - - 2.0 - 0.0

3.0 - - 3.0 - -
3.0 - - 3.0 - 0.0
3.0 - - 3.0 0.1 0.0
4.0 - 0.0 4.0 - 0.1
4.0 0.1 0.0 4.0 0.1 0.0
4.0 - 0.0 4.0 - 0.0
4.4 - - 4.4 - 0.0
6.0 - - 6.0 0.0 0.0

7.9 - - 7.9 - 0.0
8.0 - - 8.0 - 0.0
8.0 - - 8.0 0.1 0.0
9.8 - 0.0 9.8 - 0.1
10.0 0.1 0.0 10.0 0.1 0.0
12.0 - 0.0 12.0 - 0.0
12.0 - 0.2 12.0 - -
16.0 0.5 0.2 14.0 - 0.1

18.0 - 0.2 16.0 0.3 0.1
23.0 - - 18.0 - 0.2
24.0 - 0.0 20.0 0.3 0.1
23.0 - 0.1
24.0 - -

Table 4.3 - Steel pipe: Full bore hole fraction extracted from HCRD

Equipment fraction, floating Equipment fraction, floating
fraction, Steel pipe, fraction, Steel pipe,
dimension (inch) average HCR, Steel dimension (inch) average HCR, Steel
HCR 2001-2015 HCR 1992-Q12015
pipe 01-15 pipe 92-15
0.1 1.0 0.1 1.0
0.2 - 0.5 0.2 - 0.4

0.3 0.4 0.1 0.3 0.3 0.1
0.4 - 0.1 0.4 - 0.1
0.4 - 0.1 0.4 - 0.1
0.5 0.3 0.2 0.5 0.3 0.3
0.5 0.4 0.2 0.5 0.5 0.3
0.7 - 0.1 0.7 - 0.2
0.7 - 0.1 0.7 - 0.1
0.8 0.2 0.2 0.8 0.4 0.5
1.0 0.3 0.2 0.8 1.0 0.6

1.0 0.1 0.1 1.0 0.3 0.5
1.4 - 0.0 1.0 0.1 0.1
1.5 - - 1.4 - 0.1

2.0 - 0.0 1.5 0.1 0.0
2.0 0.1 0.0 2.0 - 0.1
2.0 - 0.0 2.0 0.1 0.0
3.0 - - 2.0 - 0.0
4.0 - - 3.0 0.0 0.0

5.1 - - 3.9 - 0.0
5.1 - 0.0 4.0 0.0 0.0
6.0 0.0 0.0 5.1 - 0.0
6.0 - 0.0 5.1 - -
8.0 - - 5.9 - 0.0
10.0 - - 6.0 0.1 0.0
12.0 - - 6.0 - 0.0
14.0 - - 8.0 0.1 0.1

16.0 - - 10.0 0.1 0.1
18.0 - - 12.0 - 0.0
20.0 - - 14.0 - -
30.0 - - 15.7 - -
32.0 - - 16.0 - -
36.0 - 18.0 - -
20.0 - -
30.0 - -
32.0 - -
36.0 -

Appendix C
Targets for PLOFAM model

1 Introduction .................................................................................................................................. C1
2 Targets for PLOFAM model ........................................................................................................... C2
Report no: 105586/TN-3 Rev: Final B Page Ci

1 Introduction
This Appendix contains the targets for the PLOFAM model for the equipment types with few
associated leaks.

2 Targets for PLOFAM model
Table 2.1 - Data basis for targets for equipment with few associated leaks in HCRD and few leaks
observed at installations on the NCS
Significant Leak frequency (per year)

Marginal
Dataset and Small Medium Large Maximum rate

Equipment Equipmen P-value
period (kg/s)
t years
(< 10 kg and/or Marginal Significant Sum
0.1-1 kg/s 1-10 kg/s > 10 kg/s
< 0.1 kg/s)
UKCS <2001 12 3 - 1 24.0 3 204 3.7E-03 1.2E-03 5.0E-03 0.75
Centrifugal UKCS >2001 33 3 - - 0.8 6 020 5.5E-03 5.0E-04 6.0E-03 0.25

compressor NCS >2001 - 2 - - 0.7 3 621 0.0E+00 3.0E-03 3.0E-03
Total 45 8 - 1 12 845 3.5E-03 7.0E-04 4.2E-03
UKCS <2001 16 4 - - 0.3 243 6.58E-02 1.65E-02 8.23E-02 0.75
Reciprocating UKCS >2001 34 5 - - 0.6 524 6.49E-02 9.54E-03 7.44E-02 0.25

compressor NCS >2001 - - - 1 20.0 224 0.00E+00 4.47E-03 4.47E-03
Total 50 9 - 1 991 5.0E-02 1.0E-02 6.1E-02
UKCS <2001 25 17 6 - 5.9 4 369 5.72E-03 5.26E-03 1.10E-02 0.75
UKCS >2001 59 22 4 - 1.9 7 763 7.60E-03 3.35E-03 1.09E-02 0.25

Centrifugal pump
NCS >2001 - 2 - - 0.1 3 702 0.00E+00 5.40E-04 5.40E-04
Total 84 41 10 - 15 834 5.3E-03 3.2E-03 8.5E-03
UKCS <2001 1 3 1 1 371.0 851 1.18E-03 5.88E-03 7.05E-03 0.75
UKCS >2001 8 2 2 - 7.9 1 422 5.63E-03 2.81E-03 8.44E-03 0.25

Reciprocating pump
NCS >2001 - - - - 121 0.00E+00 0.00E+00 0.00E+00
Total 9 5 3 1 2 394 3.8E-03 3.8E-03 7.5E-03
UKCS <2001 7 4 - - 0.9 1 712 4.09E-03 2.34E-03 6.43E-03 0.75
Shell side heat UKCS >2001 15 - - - - 2 921 5.14E-03 0.00E+00 5.14E-03 0.25
exchanger NCS >2001 - - - - 2 570 0.00E+00 0.00E+00 0.00E+00
Total 22 4 - - 7 203 3.1E-03 5.6E-04 3.6E-03
UKCS <2001 10 1 1 - 1.8 3 225 3.10E-03 6.20E-04 3.72E-03 0.75
Tube side heat UKCS >2001 13 2 - 1 135.0 5 834 2.23E-03 5.14E-04 2.74E-03 0.25
Total 23 3 1 1 12 537 1.8E-03 4.0E-04 2.2E-03
UKCS <2001 6 8 2 1 40.0 1 525 3.93E-03 7.21E-03 1.11E-02 0.75
Plate heat UKCS >2001 17 5 4 - 3.7 3 133 5.43E-03 2.87E-03 8.30E-03 0.25
Total 23 13 6 1 6 788 3.4E-03 2.9E-03 6.3E-03
UKCS <2001 17 5 1 - 1.8 4 937 3.44E-03 1.22E-03 4.66E-03 0.75
Shell and tube heat UKCS >2001 28 2 - 1 135.0 8 755 3.20E-03 3.43E-04 3.54E-03 0.25
exchanger NCS >2001 - - - - - 6 048 0.00E+00 0.00E+00 0.00E+00
Total 45 7 1 1 19 740 2.3E-03 4.6E-04 2.7E-03
UKCS <2001 30 7 3 1 15.3 10 437 2.87E-03 1.05E-03 3.93E-03 0.75
UKCS >2001 24 6 - - 0.4 18 898 1.27E-03 3.17E-04 1.59E-03 0.25

Process vessels
NCS >2001* - 2 - - 0.5 10 637 0.00E+00 1.88E-04 1.88E-04
Total 54 15 3 1 39 972 1.4E-03 4.8E-04 1.8E-03
UKCS <2001 17 6 1 - 15.3 4 488 3.79E-03 1.56E-03 5.35E-03 0.75
UKCS >2001 19 6 2 - 0.4 8 043 2.36E-03 9.95E-04 3.36E-03 0.25

Filters
NCS >2001* - 2 1 - 3.0 2 284 0.00E+00 1.31E-03 1.31E-03
Total 36 14 4 - 14 815 2.4E-03 1.2E-03 3.6E-03
UKCS <2001 11 2 3 3 27.8 2 084 5.28E-03 3.84E-03 9.12E-03 0.75
UKCS >2001 20 7 1 - 3.0 3 875 5.16E-03 2.06E-03 7.23E-03 0.25

Pig trap
NCS >2001* - 1 - - 0.1 786 0.00E+00 1.27E-03 1.27E-03
Total 31 10 4 3 6 745 4.6E-03 2.5E-03 7.1E-03

Table 2.2 - Targets for equipment with few incidents setting the parameters for the PLOFAM
model
Target model
Equipment Leak frequency > 0.1 kg/s per equipment year Comment
Target > Target > 1 Target >

0.1 kg/s kg/s 10 kg/s
Upper frequency limit in case of few NCS incidents 1.1E-03 Threshold target
Pooled data for UKCS and NCS for period after 2001 used.
Centrifugal Lower frequency limit in case of few NCS incidents 4.8E-04 3.92 0.69 Not known whether the 2 incidents on NCS is associated with centrifugal or
compressor Frequency estimate based on data 5.2E-04 Estimated target reciprocating compressors. Centrifugal compressor is assumed based on
population and UKCS data.
Case 2 1.88 0.21
Reciprocating Lower frequency limit in case of few NCS incidents 4.3E-03 2.69 1.68
Pooled data for UKCS and NCS for period after 2001 used.
compressor Frequency estimate based on data 8.0E-03 Estimated target
Case 2 1.79 0.18

Lower frequency limit in case of few NCS incidents 4.7E-04 3.92 0.69 Significant difference between NCS and UKCS. NCS data used (according to p-
Centrifugal pump value). Not known whether the 2 incidents on NCS is associated with
Frequency estimate based on data 2.4E-03 Estimated target centrifugal or reciprocating pumps.
Case 2 3.92 0.01
Lower frequency limit in case of few NCS incidents 2.4E-03 1.39 0.69 Pooled data for UKCS and NCS for period 1992-2014 used.
Reciprocating pump Not known whether the 2 incidents on NCS is associated with centrifugal or
Frequency estimate based on data 3.8E-03 Estimated target reciprocating pumps. Centrifugal is assumed based on population.
Case 3 0.46 0.05
Upper frequency limit in case of few NCS incidents 5.4E-04 Threshold target Amalgameted model for Shell and tube heat exhanger concluded instead of
having separate models. Total failure frequency and fraction large leaks
Shell side heat Lower frequency limit in case of few NCS incidents 1.1E-04 1.39 0.69 appear to be similar. Failure modes may be specific to shell side and tube
exchanger Frequency estimate based on data 5.6E-04 Estimated target side, but not investigated further (recommended for future projects by
execution of FMEA). Combined model more aligned with NCS data (zero
Case 3 1.39 0.17 observed incidents).
Upper frequency limit in case of few NCS incidents 4.0E-04 Threshold target Amalgameted model for Shell and tube heat exhanger concluded instead of
having separate models. Total failure frequency and fraction large leaks
Tube side heat Lower frequency limit in case of few NCS incidents 8.3E-05 1.39 0.69 appear to be similar. Failure modes may be specific to shell side and tube
exchanger Frequency estimate based on data 4.0E-04 Estimated target side, but not investigated further (recommended for future projects by
execution of FMEA). Combined model more aligned with NCS data (zero
Case 3 1.39 0.28 observed incidents).
Plate heat Lower frequency limit in case of few NCS incidents 1.4E-04 1.39 0.69 Significant difference between NCS and UKCS. NCS data used (according to p-
exchanger Frequency estimate based on data 1.7E-03 Estimated target value).
Case 2 1.39 0.07

Amalgamated model for Shell and tube heat exchanger. Significant
Shell and tube heat Lower frequency limit in case of few NCS incidents 4.8E-05 1.39 0.69 difference between NCS and UKCS. NCS data used (according to p-value).
exchanger Frequency estimate based on data 4.6E-04 Estimated target
This model is used for both Shell side and Tube side.
Case 3 1.39 0.15
Lower frequency limit in case of few NCS incidents 1.6E-04 3.92 0.69 Pooled data for UKCS and NCS for period after 2001 used.
Process vessels
Frequency estimate based on data 2.7E-04 Estimated target
Case 2 2.88 0.15
Filters
Case 2 2.54 0.04

Pig trap
Case 2 1.52 0.27


PLOFAM Report

Hochgeladen von

Dokumentinformationen

Copyright

Verfügbare Formate

Dieses Dokument teilen

Dokument teilen oder einbetten

Freigabeoptionen

Stufen Sie dieses Dokument als nützlich ein?

Sind diese Inhalte unangemessen?

Copyright:

Verfügbare Formate

PLOFAM Report

Hochgeladen von

Copyright:

Verfügbare Formate

Working together

for a safer world

Process leak for offshore

Report no: 105586/R1 Rev: Final B

Security classification of this report: Open distribution

Report no: Revision: Report date:

Prepared by: Reviewed by: Approved by:

Contributing contractor: Contributing contractor: Contributing contractor:

Jon Andreas Hestad/Geir Drage

Entity name and address: Client name and address:

Our contact: Client contact:

Report no: 105586/R1 Rev: Final B Page i

Draft 09.11.2015 First version Ingar Fossan

Report no: 105586/R1 Rev: Final B Page iii

Report no: 105586/R1 Rev: Final B Page iv

Table 1.2- Abbreviations used in main report and technical notes

Report no: 105586/R1 Rev: Final B Page v

Report no: 105586/R1 Rev: Final B Page vi

Report no: 105586/R1 Rev: Final B Page vii

1.1 Report structure

2 Leak scenarios covered by the model

Report no: 105586/R1 Rev: Final B Page 1

Report no: 105586/R1 Rev: Final B Page 2

Table 2.2 - Leak scenarios suggested for QRAs at NCS

Modelled leak scenario Leak scenarios included

Report no: 105586/R1 Rev: Final B Page 3

Report no: 105586/R1 Rev: Final B Page 4

3.1 Model summary

Report no: 105586/R1 Rev: Final B Page 5

(, ) Hole size frequency distribution (see TN-1) [year-1 component-1].

0 Total leak frequency [year-1 component-1].

The total full bore hole frequency [year-1 component-1]

Hole size diameter [mm]

Equipment diameter [mm]

The average leak frequency (independent of equipment diameter) for the

0 Parameter in equation for total hole frequency

0 Parameter in equation for total hole frequency

Parameter in equation for full bore hole frequency

Parameter in equation for full bore hole frequency

Parameter in equation for full bore hole frequency

() Adjustment factor of the total leak frequency relative to the weighted

() Full bore hole fraction

Report no: 105586/R1 Rev: Final B Page 6

4 Equipment types included in the model

Table 4.1 - Equipment types included in the model

Equipment type Additional description

Report no: 105586/R1 Rev: Final B Page 7

5.1 NCS data

Report no: 105586/R1 Rev: Final B Page 8

5.2 UKCS data

6 Model parameterisation and validation

Report no: 105586/R1 Rev: Final B Page 9

Figure 6.1 - Main steps in parameterisation process

Table 6.1 - Equipment types subjected to separate parameterisation process

Hoses No data available in HCRD on leaks occurring under temporary operations

The validation model has demonstrated that PLOFAM is able to:

Report no: 105586/R1 Rev: Final B Page 10

Report no: 105586/R1 Rev: Final B Page 11

Air-cooled heat exchanger 1.00 - - - 2.70E-02 5.67E-04 -

Report no: 105586/R1 Rev: Final B Page 12

Report no: 105586/R1 Rev: Final B Page 13

Report no: 105586/R1 Rev: Final B Page 14

Report no: 105586/R1 Rev: Final B Page 15

Report no: 105586/R1 Rev: Final B Page 16