Professional Documents
Culture Documents
Guide To Finding and Using Reliability Data For QRA: Source Responsible
Guide To Finding and Using Reliability Data For QRA: Source Responsible
434-20 2019
responsible source
Acknowledgements
Safety Committee
Feedback
Disclaimer
Whilst every effort has been made to ensure the accuracy of the information
contained in this publication, neither IOGP nor any of its Members past present or
future warrants its accuracy or will, regardless of its or their negligence, assume
liability for any foreseeable or unforeseeable use made thereof, which liability is
hereby excluded. Consequently, such use is at the recipient’s own risk on the basis
that any use by the recipient constitutes agreement to the terms of this disclaimer.
The recipient is obliged to inform any subsequent recipient of such terms.
This publication is made available for information purposes and solely for the private
use of the user. IOGP will not directly or indirectly endorse, approve or accredit the
content of any course, event or otherwise where this publication will be reproduced.
Copyright notice
The contents of these pages are © International Association of Oil & Gas Producers.
Permission is given to reproduce this report in whole or in part provided (i) that
the copyright of IOGP and (ii) the sources are acknowledged. All other rights are
reserved. Any other use requires the prior written permission of IOGP.
Revision history
Contents
Abbreviations 5
7. References 41
4
Guide to finding and using reliability data for QRA
Abbreviations
5
Guide to finding and using reliability data for QRA
1.1 Scope
The reliabilities of Instrumented Protection Systems (IPFs) is a key input to Quantitative
Risk Assessment (QRA) of hydrocarbon exploration and production facilities. The IPFs
include systems such as the Fire and Gas System (FGS), Emergency Shutdown (ESD) and
blowdown systems, High Integrity Pressure Protection System (HIPPS), Process Shutdown
(PSD) systems and blowout prevention.
This datasheet provides guidance on obtaining, selecting and using reliability data for these
systems and for their component parts, for use in QRA.
1.2 Application
This datasheet contains specimen data taken from previous OGP datasheets; this specimen
data is presented in the appendices. In addition, the recommended data sources that are
identified in section 2.2 should be consulted to ensure that all data are the most up to date
and relevant for any analysis. Guidance on using and processing data is given in Section 3.
The data presented are applicable to activities in support of operations within exploration
for and production of hydrocarbons.
1.3 Definitions
For the purposes of this document, the following terms and definitions apply.
Common Cause Failures Concurrent failures of different devices, resulting from a single
event, where these failures are not consequences of each
other. These failures may occur at the same time or within a
short time of each other.
6
Guide to finding and using reliability data for QRA
7
Guide to finding and using reliability data for QRA
Mean Repair Time (MRT) Expected overall repair time. This includes the following:
• Time spent before performing a repair (after a fault has
been detected); i.e., acquiring resources, and/or spare
parts, including potential logistic delays, and taking the
system out of service, including time required for insulation
removal, scaffolding preparation, warm up/cool down &
drying time;
• Time spent to perform the repair; and
• Time spent to put the repaired component back into
operation, i.e., time to restore all the necessary insulation,
warm up/cool down & dry out, test the repaired component,
scaffolding removal (if intrusive), etc.
Mean Time Between Predicted time between failures of a system during operation.
Failures (MTBF) The MTBF assumes that the failed system is immediately
repaired (mean time to repair, or MTTR), as a part of a renewal
process. This contrasts with the mean time to failure (MTTF),
which measures average time to failure if the system is not
repaired (infinite repair time).
8
Guide to finding and using reliability data for QRA
9
Guide to finding and using reliability data for QRA
Database
Categories
Electrical &
Mechanical Human
Machinery Electronic
Parts Reliability
Parts
Non-
Rotating Electrical Electronic Process
General Electronic Valves
Machines Components Parts Vessels
Parts
OREDA Handbook 2015, 6th Edition – Topsides Equipment OREDA Participants http://www.oreda.com/
Volume 1
OREDA Handbook 2015, 6th Edition – Subsea Equipment OREDA Participants http://www.oreda.com/
Volume 2
10
Guide to finding and using reliability data for QRA
Nonelectronic Parts Reliability Data – Electrical Quanterion Solutions RMQSI Knowledge Centre
NPRD-2016 Assemblies and Incorporated/RIAC
Electromechanical/
Mechanical Parts
Electronic Parts Reliability Data – EPRD- Electronic components Quanterion Solutions RMQSI Knowledge Centre
2014 Incorporated/RIAC
Failure Mode/Mechanism Distributions Electrical, mechanical, Quanterion Solutions RMQSI Knowledge Centre
FMD-2016 and electromechanical Incorporated/RIAC
parts
PDS Data Handbook, Reliability Data for Sensors, Detectors, SINTEF PDS
SIS, 2013 Edition Valves and Logic https://www.sintef.no/
Solvers projectweb/pds-main-page/
493-2007 - IEEE Recommended Practice Electrical power IEEE Standards Association http://www.techstreet.com/
for the Design of Reliable Industrial and generation and ieee
Commercial Power Systems distribution equipment
Kirwan, B. (1994) A Guide to Practical Human factors CPC Press CPC Press
Human Reliability Assessment. CPC
Press.
The validation of three human reliability Human factors Applied Ergonomics. 27(6) Applied Ergonomics. 27(6)
quantification techniques - THERP, 359-373 359-373
HEART, JHEDI
Part 1, 2 and 3
Note: Some of the data sources listed in Table 2-1 are now considered to be outdated, especially for electronic components as
the technology has advanced significantly since these were compiled. For example, the MIL-HDBK-214F [1] is over 20 years old.
Use of these should be limited to specific cases when contemporary data sources do not contain the information.
11
Guide to finding and using reliability data for QRA
Pressure
Electronic Electrical Logic Process Subsea
Data Source Sensors Valves Safety
Components Machines Devices Vessels Equipment
Devices
OREDA Handbook Vol. 1 Yes Yes Yes Yes Yes Yes Yes
NPRD-2016 Yes
12
Guide to finding and using reliability data for QRA
3.1 Introduction
Reliability is a broad term that covers multiple aspects of a system or product and focuses
on the ability of a product to perform its intended function. Mathematically, if an equipment
item is performing its intended function at time equals zero, reliability can be defined as
“the probability that an item will continue to perform its intended function without failure
for a specified period under stated conditions”. The product defined here could be an
electronic or mechanical hardware product, a software product, a manufacturing process
or even a service.
The science of reliability prediction is based upon the principals of statistical analysis.
Reliability engineering uses a probabilistic approach rather than a deterministic one. This
probability can be calculated or stated to reside within certain statistical confidence limits.
Fundamental to such a calculation is the ability to source basic reliability data. Ideally such
data1 should be:
• Current
• Auditable
• Specific (applicable to equipment/component type)
• Extensive (large sample with many recorded failures)
• Applicable to environment
• Be suitable for life trending
Unfortunately, real world data sources rarely meet these ideals and it is therefore
necessary to accept compromises. When performing QRA, it is important that the
limitations of the data source are understood, and where necessary alternatives sought.
For QRA, the reliability parameters to be taken from the database would be the failure rate
(or the mean time to failure, MTTF) and/or the average probability of failure on demand
(PFDavg). See Section 3.3 for details of probability of failure on demand calculation.
Where information is extracted from the OREDA [2] or another industry standard
database it is not (in general) necessary to perform any further statistical analysis of the
failure patterns for QRA purposes. The approach described in Section 3.2 applies where
basic information relating to times to failure is available for analysis, for example from
maintenance records or breakdown reports. In these circumstances, it is necessary to
judge the quality of the data and to then apply the appropriate analytical technique. The
techniques for data analysis presented herein are divided into two classifications as follows:
• Based on sample statistics
• Based on inferences from the associated statistical distributions
13
Guide to finding and using reliability data for QRA
The characteristics of distributions are much harder to derive, especially from field
breakdown reports rather than laboratory test data, but have the potential to provide
more information. Note that it is not the intention to provide a comprehensive theoretical
background to data analysis in this document, but instead to provide some practical
techniques that may be used to prepare reliability data.
3.2.1 Background
The observed failure rate for a component is defined as the ratio of the total number of
failures to the total cumulative observation or operational time. For items displaying a
constant failure rate, if λ is the failure rate of the N items then:
λ=k⁄T
Where k is the total number of failures and T is the total observation time across the N items.
For the case where components are replaced after failure (as applies to industry field
databases), then the total cumulative observation time may be defined as N × field
operational lifetime.
Strictly, this calculation provides a point estimate of the failure rate and if the exercise were
repeated with another set of identical equipment and conditions, it may yield results that
are not identical to the first. Any number of such measurements may be made providing a
number of “point estimates” for the failure rate, with the true value of the failure rate only
being provided after all components have failed (for a non-replacement test). In practice,
therefore, it is necessary to make a prediction about the total population of items based
on the failure patterns of a sample. This process of statistical inference can be performed
using the properties of a “χ2” (chi squared) distribution. This allows us to bound the
population failure rate within confidence limits (typically 90% or 60% may be used).
It is also necessary to make some assumptions about the pattern of failures across time,
considering the shape of the commonly depicted ‘bathtub curve’ (Figure 3.1). This curve
typifies the expected component failure rate across time and is divided into three distinct
areas, namely:
• Early life: Characterized by a decreasing failure rate
• Useful life: Constant failure rate
• Wear out: Increasing failure rate
14
Guide to finding and using reliability data for QRA
To perform analysis of failure patterns outside of the constant failure rate period a level of
detailed information is required that is typically not available from the recorded data (e.g.,
actual age of equipment of failure, homogeneous samples). Therefore, an assumption
is made that all failures recorded are experienced during the useful life phase, and the
pattern of these failures may be described by a random, exponential distribution. This can,
at least to a certain extent, be justified on the following grounds:
• Early life failures resulting from commissioning problems may not be recorded as
equipment failures
• Early life failures resulting from manufacturing defects can be largely eliminated by
testing prior to installation
• Wear out failures largely eliminated by preventative maintenance and planned
renewals. Note that this assumption may be less valid for wear out of subsea
equipment where no planned maintenance will be performed
The discussion allows us to analyse the data from each source, and in most cases to
calculate a mean value, confidence intervals about the mean value and the associated
variance.
Of these, the constant failure rate (point estimate or Chi-Squared failure rate) model is
most commonly used for QRA purposes.
15
Guide to finding and using reliability data for QRA
If k is the total number of failures of N items, then the failure rate λ(t) is given by:
λ(t) = k ⁄ T
where T is the total cumulative observed time.
As the product matures, the weaker units die off, the failure rate becomes nearly constant,
and modules have entered what is considered the normal life period. This period is
characterised by a relatively constant failure rate. The length of this period is referred to
as the system life of a product or component. It is during this period that the lowest failure
rate occurs. The amplitude on the bathtub curve is at its lowest during this time. The useful
life period is the most common time frame for making reliability predictions. Most of the
failure rates quoted in data references (such as MIL-HDBK-217) apply to this period.
Also, during the “constant failure rate” period, the mean time between failures (MTBF) is
often reported instead of the failure rate. MTBF can be obtained by:
MTBF = 1⁄λ
Note: This assumption is only valid for the flat region of the bathtub curve. It is inappropriate
to extrapolate MTBF to give an estimate of the service life time, which may be less than
suggested by the MTBF given the higher failure rates in the wear-out part of the bathtub curve.
The MTBF numbers are preferred in engineering usage as large positive numbers (say 5000
hours) is more intuitive than very small numbers (say 0.0002 per hour) and can be linked to
maintenance intervals based on hours of operation.
This can be addressed via a Chi Squared (χ2) test using the methodology described below.
This method can forecast the failure rate when no failures have been recorded in the
observed time using confidence intervals.
16
Guide to finding and using reliability data for QRA
Calculation
1) Calculate α = 1 - CI
2) Calculate n as follows:
Single-Sided Limits*
a) n = 2k for failure-truncated test
b)
n = 2(k+1) for time-truncated test
Look up the value of χ2 corresponding to n and α (using statistics tables for Chi Squared
distribution or see Table 3-1.
3) Failure Rate Confidence Limit λCI is calculated using λCI = (χ2) ⁄ 2T
* Confidence Limits
The limits defining the interval are called confidence limits. These are the highest and the
lowest values in the interval. The two-sided version tests against the alternative that the
true variance is either less than or greater than the specified value. The one-sided version
only tests in one direction. The choice of a two-sided or one-sided test is determined by the
problem.
𝑓𝑓(𝛘𝛘𝟐𝟐) 𝑓𝑓(𝛘𝛘𝟐𝟐)
1 − 𝛼𝛼 𝛼𝛼 𝛼𝛼)
2 1 − 𝛼𝛼 𝛼𝛼)
2
𝛘𝛘𝟐𝟐 𝛘𝛘𝟐𝟐
It is worth noting that λCI (χ2 ⁄ (2T) is a conservative estimate. The true value has probability
of α being higher than the estimate (based on a single sided upper confidence limit). Using
the upper bound of the failure rate is a conservative approach and hence it can be used
instead of the maximum likelihood estimate when the sample is small.
17
Guide to finding and using reliability data for QRA
Worked Example
Problem: Equipment maintenance records show that 5 identical devices each with a
recorded running time of 1000 hours each have experienced no recorded failures. Calculate
the failure rate at 90% confidence (single sided upper limit).
Solution:
18
Guide to finding and using reliability data for QRA
Note: The decision to use statistical interpretation or point estimate is based on the
number of recorded failures. For items with a very high failure rate, a significant number of
failures could equate to a small amount of experience years, but typically a large amount of
experience years is also required for a point estimate.
There are special cases associated with values of the shape factor:
• β=1 corresponds to exponential distribution
• β<1 represents burn in (decreasing failure rate)
• β>1 represents wear out (increasing failure rate)
Note: In line with convention, β is used here to represent the shape factor of the Weibull
distribution. This is not the same β used to describe the dependent failure fraction of
common cause failures.
19
Guide to finding and using reliability data for QRA
By using a graphical plotting technique, the data can be quickly analysed without detailed
knowledge of statistical mathematics. A simple procedure for this is as follows:
• Determine test sample size and times to failure
• List times to failure in ascending order
• Establish median rankings from published tables (or calculate/estimate from
formulae)
• Plot times and corresponding ranks on Weibull plot paper. This is essentially log-log
graph paper but with scales for reading β and α
• Draw best fit straight line and read off α at 63.3% intercept
• Draw a parallel line through intercept on y axis and read off β
Median ranking is the most frequently used method for probability plotting, especially if the
data are known not to be normally distributed. Median ranking tables are available from
statistics text books, or they may be estimated by the following equation:
R = (i - 0.3) / (N + 0.4)
where i is the failure order number and N is the total number of failures.
In the following example, failures are listed from 1 to 10 with their corresponding time to
failure and median rank. These are then plotted on the Weibull paper.
20
Guide to finding and using reliability data for QRA
Plot Line and Read Values of characteristic life (α) and shape factor (β)
It is generally acceptable to fit a straight-line plot by eye through the data points. The value
of shape factor is read by drawing a line perpendicular to the plotted line through the plot
origin. The value of β can then be read from the intercept of this line and the β scale. The
value for the characteristic life may be read from the intercept of the plotted line with
the “estimator line”. The position of the estimator is determined by the intercept of the
perpendicular line with the α scale.
In the above plot all three stages of the bathtub curve are displayed, the values are
approximately:
21
Guide to finding and using reliability data for QRA
(n + 1) - iti-1
where Nti =
1 + (n - number of preceding items)
The third Weibull parameter (location parameter), γ, locates the distribution along the
abscissa. Changing the value of γ has the effect of “sliding” the distribution and its
associated function either to the right (if γ > 0) or to the left (if γ < 0). The parameter γ may
assume all values and provides an estimate of the earliest time a failure may be observed.
A negative γ may indicate that failures have occurred prior to the beginning of the test or
prior to actual use. The life period 0 to +γ is the failure free operating period of such units.
To account for this, an attempt can be made to predict the failure free period. This may
be based on engineering judgement and knowledge of the items under consideration or
may simply be the time until the first failure occurs. The data are then replotted from this
time and if a straight line results the failure free period is as estimated and the remaining
parameters may be estimated from the plot. If another curve is produced the process is
repeated.
22
Guide to finding and using reliability data for QRA
Several mathematical techniques exist for the treatment of CCF’s. One of the simplest
and most practical is the Beta factor approach. This assumes that λ, the total failure rate
for each redundant unit in the system, is composed of independent and dependent failure
contributions as follows:
λ = λc + λi
where λi is the failure rate for independent failures
λc the failure rate for dependent failures
The parameter beta ( β) can then be defined as:
β = λc /λ
Note: β is also commonly used to represent the shape factor of the Weibull distribution,
this is not the same as β used to describe the dependent failure fraction of common cause
failures.
Thus, β is the relative contribution of dependent failures to total failures for the item. The
lack of available data relating to dependent failures of sufficient quality necessitates the
use of an estimation technique for beta, guided by several parameters shaping factors (the
subjective assessment of defensive mechanisms). Such a quantification method, known as
the partial beta factor model may be applied for detailed assessment.
For a simpler approach a representative value of β may be assumed between 0.01 (highly
diverse components or systems) and 0.1 (similar components or systems).
The handbook also gives point estimates of failure rate; the numerical difference between
this and the OREDA estimator gives an indication of the degree of diversity in failure rates
between parts of the overall population. OREDA recommends that the OREDA estimator be
used when data are taken from this source.
23
Guide to finding and using reliability data for QRA
For a protective system having dangerous failure rate λd and proof test interval T, the
probability of failure on demand or unavailability due to unrevealed failures is presented in
Table 3-3.
The table gives a simplified set of equations to calculate the PFDavg for different redundant
architecture combinations usually represented by MooN (M out of N). For example, an on-
demand 2oo3 system implies that 2 elements must work correctly out of 3 for the overall
system to work successfully.
Further details are available in functional safety standard, IEC 61508 Part 6 [5].
PFDavg λd T
⁄2 λ2 T 2
d ⁄3 λ T
d
λ2 T 2
d
λ3 T 3
d
Note that these simplified formulae are only applicable under specific conditions. For
example, 1oo1 must have λdT ≪ 1.
24
Guide to finding and using reliability data for QRA
Operational Mode Is the data appropriate for the operating mode of the equipment?
Equipment operated frequently in a standby mode (emergency
generators, firewater pumps) will exhibit different failure modes
and frequency compared to equipment operating continuously.
Number of Recorded Is the equipment failure data set large enough to be
Failures representative?
Equipment with few recorded failures will have a large uncertainty
associated with their failure rate. In such cases, Chi-Squared
method may be better.
Population and Does the data set encompass wide a enough population?
Installations It is desirable for data to be selected for equipment with a large
population across a wide number of installations. This avoids
data representing localised effects or dominated by one design or
manufacturer.
Time in Service Has the data being gathered from equipment which has spent
sufficiently long time in similar conditions?
It is desirable for data to be selected for equipment with a long
time in service (calendar time). The operational time may be
considerably less for equipment that is normally on standby (e.g.,
firewater pumps).
25
Guide to finding and using reliability data for QRA
λCritical
Healthy Critical
Non-
λDegradation Degradation
Critical
λIncipient
Incipient
In some cases, OREDA does not have any values recorded against ‘Critical’ failure class.
In such cases, it can be assumed that a proportion of degradation and incipient failures
can result in critical failure if not suitably attended to in good time. For simplicity, the
QRA analyst can assume that critical failure rate is the summation of the degradation and
incipient failure rates. However, the following points must be true for this to be a valid
assumption:
• The sample size of the data set is sufficiently large
• No credit has been taken for preventative maintenance
The QRA analyst could also consider other data sources which may have the requisite
data. If weighting of data between different sources is considered necessary, the method is
available in Estimation Procedures section of OREDA [2].
Although unlikely, an analyst may be required to use more sophisticated methods. Such
methods are available, for example Bayesian estimation techniques, where the parameters’
prior distributions are founded on a broader range of the data gathered within in the OREDA
project [31].
26
Guide to finding and using reliability data for QRA
Note: Access to the electronic database is restricted to participants in the OREDA program.
27
Guide to finding and using reliability data for QRA
OREDA data equipment groups and the equipment items covered are listed in Table 4-2.
Volume 1
1.3 Pumps
1.5 Turboexpanders
3.2 Vessels
Volume 2
5.2 Flowlines
5.3 Manifolds
5.5 Risers
5.7 Templates
Boundaries
Each equipment item class has an inventory description provided at the start of the
respective chapter. This should be examined carefully to identify equipment items for the
28
Guide to finding and using reliability data for QRA
system under consideration that lie outside the defined OREDA boundary. These must then
be considered as separate items. An example of this would be a compressor or electrical
generator where the prime mover is listed as a separate item.
Taxonomy code
The taxonomy code gives an identification of the equipment item selected from the
database. It is good practice to record this code and to include it within calculations as a
reference for any data extracted.
Population
Total recorded observation time for the population when it is required to fulfil its functional
role. Note that this may be an estimated value.
Number of demands
Total number of recorded demand cycles for the population. Note that this may be an
estimated value.
Failure Mode
This column presents the recorded modes of failure for the equipment item, divided into
severity classes critical, degraded, incipient and unknown. In general, only the critical
severity class failures need be considered, i.e., those that cause an immediate and
complete loss of an items function. Where an equipment item performs more than one
function (e.g., process and protective) it may be necessary to review each failure mode and
identify the requirement to progress it into the risk calculation, either as an aggregated
failure rate value for the equipment item or as individual failure events. i.e., critical failures
may include dangerous, non-dangerous and safe failures. These failures may be critical to
production but not to the equipment’s protective function.
Number of Failures
This is the total number of failures aggregated across all modes. In general, the higher the
number of failures, the greater the confidence in the calculated failure rate.
Failure Rate
All failure rates in the OREDA handbook are presented in terms of failures per million
hours. The following data are presented for each mode, calculated both in terms of
calendar and operational time:
29
Guide to finding and using reliability data for QRA
• Mean: estimated average failure rate, calculated using the “OREDA” estimator – see
Section 3.2.6 for details
• Lower, Upper: 90% confidence bounds for the failure rate
• SD: Standard deviation
• n/T: Point estimate of the failure rate i.e. total number of failures divided by the total
time in service
For most calculations, it is recommended that the mean value (i.e., based on the OREDA
estimator) is used. Note that the difference in value between the point estimate and mean
failure rate relates to the degree of diversity in the population.
4.2 MIL-HDBK-217F
The MIL-HDBK-217 [1] handbook contains failure rate models for the various part types
used in electronic systems, such as integrated circuits, transistors, diodes, resistors,
capacitors, relays, switches, and connectors.
The handbook details two methods for reliability prediction, namely parts count and parts
stress calculation. Parts count prediction is recommended during the design phase of a
project. It is simpler than parts stress and requires less detailed information. To calculate a
system failure rate the following method is used:
For each component part of a system, a baseline failure rate value is selected from tables
based on the type of the part and the operating environment. This value is then modified
by multiplying by a quality factor, again selected from a table (e.g., military or commercial
specification). For microelectronics, a learning factor may also be applied. The overall
system failure rate is then derived by summation of the parts failure rates; hence the
title “parts count”. In general, parts count analysis will provide an adequate estimate of a
system’s failure rate for use in QRA.
Parts stress analysis involves derivation of more multiplying factors that in turn require
detailed analysis of the system.
4.3 FIDES
This is reliability standard created by FIDES Group - a consortium of leading French
international defence companies: AIRBUS, Eurocopter, Giat, MBDA and THALES. The FIDES
methodology is based on the physics of failures and is supported by the analysis of test
data, field returns and existing modelling. The FIDES Guide is a global methodology for
reliability engineering in electronics. It has two parts, namely a reliability prediction guide
and a reliability process control and audit guide.
30
Guide to finding and using reliability data for QRA
The design of reliable power distribution systems is significant because of the high cost
associated with power outages. It is necessary to consider the cost of power outages when
making design decisions for power distribution systems.
31
Guide to finding and using reliability data for QRA
A study (JIP) on reliability of well completion equipment (“Wellmaster Phase III”) [11] was
completed by SINTEF in November 1999. This has resulted in a database of well completion
equipment, with a total of 8000 well-years of completion experience represented.
A subsea equipment reliability database project was completed by ExproSoft in late 2000
(Phase I) [12]. This project, led to the development of the SubseaMaster database and
software version 1.0. Phase II of SubseaMaster was launched as a joint industry project in
May 2001. and was completed in April 2003.
Weblink: http://www.exprosoft.com/products/wellmaster-rms/
32
Guide to finding and using reliability data for QRA
In this section, reliability as aspects pertaining to various types of protection systems used in
oil and gas installations is discussed. Note that the reliability figures quoted are representative
values for the relevant system, and are subject to assumptions and limitations (such as
adequate maintenance being carried out on the components, operating conditions etc.).
Safety Systems
Emergency Quick
Disconnect Systems
High Integrity
Pressure Protection
System
Blowdown
System
33
Guide to finding and using reliability data for QRA
If the QRA analyst is aware of a system being classed as a SIS, it is better to obtain the
achieved reliability data for use in the assessment. The following table gives the typical
values that can be used in absence of such information or for a crude estimate. These
are based on geometric mean of the SIL range as suggested by UKOOA guidelines [13].
However, arithmetic averages can also be used.
SIS Type/Typical SIL PFDavg (Geometric Mean) Risk Reduction Factor (RRF)
3 In many cases, some functions within the PSD and ESD may be classified as Safety Instrumented Functions (SIFs). The overall system
that executes the SIF is the SIS. These systems have a defined range of reliability that they must meet to comply with the functional
safety standards. Hence, it is recommended to use the ‘minimum achieved reliability’ (also: “the lower bound”)’ for these systems in
the QRA process in absence of any relevant information on such systems. A SIL verification report typically will contain the achieved
reliability for these systems.
34
Guide to finding and using reliability data for QRA
Note 1: Blowdown systems and fire and gas detection systems are not SIS in the strict
sense. These systems are typically treated as being equivalent to SIL 1, i.e., these are
typically engineered to achieve at least SIL 1 reliability range. If such systems are relatively
old (i.e., nearing obsolesce or end-of-life), a reduced reliability range is recommended
to allow for normal degradation and systematic errors which may not be rectified using
maintenance. The SIL target only considers the reliability of hardware and does not reflect
the overall reliability of such systems.
Note 2: The reliability values are for an automatic blowdown system which is activated as
part of the safety logic. For manual initiation of the blowdown system, it is suggested that
60% - 70% of availability is assumed.
At a sub-system level (sensors, logic solver and final elements), the probability of failure on
demand is dependent upon the test interval as mentioned in section 3.3. An example (using the
formulae in section 3.3) depicting the impact of proof testing interval is shown in Figure 5.2.
In the example, dangerous failure rate (λd) of 8.00 x 10-3 has been assumed. The figure
shows that PFD average increases with increase in the interval between tests. There is
usually an optimal test interval for each system or function which balances the test interval
requirements against cost of conducting tests and other variables.
Impact of Proof Test Interval
1.00E+00
1.00E-01
1.00E-02
PFD Average
1.00E-03
1.00E-04
1.00E-05
Test Interval (Years)
1.00E-06
0.25 0.5 1 2 3
1oo1 1.00E-03 2.00E-03 4.00E-03 8.00E-03 1.20E-02
1oo2 1.33E-06 5.33E-06 2.13E-05 8.53E-05 1.92E-04
2oo2 2.00E-03 4.00E-03 8.00E-03 1.60E-02 2.40E-02
2oo3 4.00E-06 1.60E-05 6.40E-05 2.56E-04 5.76E-04
35
Guide to finding and using reliability data for QRA
Isolation Blowdown
Case 1 Performs as per design Performs as per design
Case 2 Performs as per design Fails
Case 3 Fails Performs as per design
Case 4 Fails Fails
The probability of each case combination can be estimated using the overall reliability of
each system. For case 3, the installation’s shutdown logic should be consulted to ensure
that it permits blowdown without isolation.
Active fire protection systems comprise water deluge systems (pumps, deluge valves, ring-
main, deluge nozzles, monitors), chemical foam systems, water mist systems etc. Passive
protection systems comprise passive fire protection (PFP) coatings, firewalls and blast
walls. For cryogenic spills, a cryogenic spill protection system may be used.
36
Guide to finding and using reliability data for QRA
For modern fire water systems, an availability of the deluge system of at least 90%
is required but 99% is desired. Various analyses show that the fire water systems
unavailability is largely dependent on the reliability of the deluge valves. The deluge valves
constitute 95.5% of the total unavailability.
For purposes of a QRA, data presented in Appendix 14 (Failure and Event Data) in Lee’s
Loss Prevention in the Process Industries: Hazard Identification [15] can be used. A
summary is presented in Table 5-4.
National Fire Protection Association (NFPA) standards are widely used for defining required
water/foam delivery rates and durations along API 6F series for fire tests and API 2218 for
fireproofing etc. The most common standards used by design engineers are listed in Table 5-5.
Code Title
NFPA 4 Standard for Integrated Fire Protection and Life Safety System Testing
NFPA 10 Standard for Portable Fire Extinguishers
NFPA 11 Standard for Low-, Medium-, and High-Expansion Foam
NFPA 12 Standard on Carbon Dioxide Extinguishing Systems
NFPA 12A Standard on Halon 1301 Fire Extinguishing Systems
NFPA 13 Standard for the Installation of Sprinkler Systems
NFPA 14 Standard for the Installation of Standpipe and Hose Systems
NFPA 15 Standard for Water Spray Fixed Systems for Fire Protection
NFPA 16 Standard for the Installation of Foam-Water Sprinkler and Foam-Water
Spray Systems
NFPA 17 Standard for Dry Chemical Extinguishing Systems
NFPA 17A Standard for Wet Chemical Extinguishing Systems
NFPA 20 Standard for the Installation of Stationary Pumps for Fire Protection
NFPA 25 Standard for the Inspection, Testing, and Maintenance of Water-Based
Fire Protection Systems
NFPA 99B Standard for Hypobaric Facilities
37
Guide to finding and using reliability data for QRA
Code Title
NFPA 130 Standard for Fixed Guideway Transit and Passenger Rail Systems
NFPA 750 Standard on Water Mist Fire Protection Systems
NFPA 770 Standard on Hybrid (Water and Inert Gas) Fire Extinguishing Systems
NFPA 1150 Standard on Foam Chemicals for Fires in Class A Fuels
NFPA 2001 Standard on Clean Agent Fire Extinguishing Systems
NFPA 2010 Standard for Fixed Aerosol Fire-Extinguishing Systems
ISO 13702:2015 Petroleum and natural gas industries – Control and mitigation of fires
and explosions on offshore production installations – Requirements and
guidelines
In addition, legacy systems may have been designed to the standard BS 5306 [16]. This
standard mandated water delivery rate of 9.81 litres/min/m2 over the exposed vessel
surface and its supports is required. For protection from lower levels of thermal radiation
from fires on adjacent units, lower rates of water application are allowable.2626
For PFP based systems, typically, the criteria will be that a protected surface will not reach
a certain temperature in a defined time-period during a standard test. The protective
system should meet the requirements of a fire tests (example: pool fire test as described in
UL 1709 [17] and jet fire test as described in ISO 22899 [18]). For well-maintained systems,
an availability of the at least 90% is required but 99% is desired.
In QRAs, the analyst should also consider DAL values for items such as firewalls and blast
walls. If the thermal radiation or overpressure from an event exceeds the specified DAL, the
protection measure should be assumed to be impaired and ineffective.
The proof test should include all parts of the overall system including “standby” equipment.
38
Guide to finding and using reliability data for QRA
Similarly, effectiveness of PFP and other passive means of safety may be severely
compromised if the hazardous event load exceeds DAL. In such cases, PFP can be
damaged or even blown-off the equipment it is protecting.
Typically, an availability of 90% to 99% on a sliding scale can be assumed for such systems
provided they meet the maintenance requirements, and DAL is not exceeded.
The reliability of evacuation crafts or totally enclosed motor propelled survival craft
(TEMPSC) (or lifeboats and life rafts) is difficult to ascertain due to lack of data. HSE
research report RR599 [19] discusses the issue in detail. However, for QRA purpose,
the assumed reliability should not exceed 90% for newer crafts. Ageing crafts undergo
degradation in the glass reinforced plastic (GRP), and hence offer a much lower reliability
as a system.
39
Guide to finding and using reliability data for QRA
Layer of Protection Analysis – Simplified Process Risk Assessment [21] also presents
worked examples together with some specimen reliability data.
Background reliability theory can be found in Practical Reliability Engineering [22] and
Reliability, Maintainability and Risk [3]. The latter also contains some reliability data from
FARADIP [4].
Reliability Technology [23] contains (older) reliability data from the nuclear industry.
40
Guide to finding and using reliability data for QRA
7. References
41
www.iogp.org
Registered Office Brussels Office Houston Office
City Tower Avenue de Tervuren 188A 19219 Katy Freeway
Level 14 B-1150 Brussels Suite 175
40 Basinghall Street Belgium Houston, TX 77094
London EC2V 5DE USA
T +32 (0)2 790 7762
United Kingdom
eu-reception@iogp.org T +1 (713) 261 0411
T +44 (0)20 3763 9700 reception@iogp.org
reception@iogp.org