You are on page 1of 7


Technical Note

Data sources for reliability design analysis

T R Moss, MPhil, CEng, FIMechE, FIQA
Reliability Consultant, Abingdon, Berkshire
J E Strutt, BSc, DPhil, MSUT
Centre for Industrial Safety and Reliability, Cranfield Institute of Technology, Bedfordshire

Quantijied reliability assessment is now employed by many organizations in the process industries. The objective is to identify critical
areas in the plant where design changes are required to meet the safety or production specification. Mechanical system assessment is,
however, frequently inhibited by the lack of representative equipment reliability data. Where these data are not available in-house
recourse must be made to generic sources from the public domain. Members of the Process Industries Divisions Mechanical Reliability
Committee have recently completed a study to identify and evaluate the principal reliability data sources for use in process system
assessments. This paper discusses the data requirements for reliability assessment and the use of these generic data to provide best-
estimates of the reliability statistics for specific applications.

1 INTRODUCTION Such an assessment should identify the major hazards

The use of quantitative reliability analysis to optimize (safety or production hazards depending on the focus of
safety and productivity of new and existing plant has the study) and the equipment failures that are most
become widespread in the process industries. Effective likely to lead to the materialization of the hazard.
use of the technique does, however, require access to Depending on the consequences of system failure it may
statistics on the failure and repair characteristics of then be necessary t o consider introducing some
equipment. Unfortunately sources of such information redundancy or diversity, of critical equipment, the
are relatively sparse in the literature and significant time reduction of hazardous material inventories or changes
must be spent to assemble a database of relevant relia- in maintenance or test procedures to improve the relia-
bility statistics for each study. bility of the system and reduce the hazards associated
To address this problem a sub-group of the IMechE with system failure.
Mechanical Reliability Committee has carried out a
number of studies over the past 5 years aimed at iden-
tifying and evaluating the main sources of generic relia- 2.2 System definition
bility data for mechanical equipment. The information Most engineering systems are complex to some degree
has now been summarized on to a series of data sheets and the reliability assessment will, therefore, need to
for publication later this year in the second edition of follow a consistent, logical approach, It is always neces-
the Process Industries Divisions guide to reliability sary to define the objectives of the assessment and the
assessment (1). functional attributes of the system at an early stage in
This paper discusses the basis for reliability studies of the study. The objectives may be safety or productivity
process systems and demonstrates how information orientated. If safety is the prime concern the assessor
from these data sources can be applied to generate best- may need to evaluate the hazards to people living near
estimate reliability statistics for specific equipment. the plant and the environment as well as to plant per-
sonnel and the process. If it is a production reliability
study, is the objective to increase plant availability gen-
erally or for peak demand, to study maintenance or test
2.1 General procedures, spares holdings, manning levels or other
aspects of plant performance? A clear definition of the
Assessing the reliability of a process plant design systems function and the objectives of the assessment is
involves an analysis of the ways a system can fail from
the best guarantee that a clear set of recommendations
consideration of its sub-system and equipment failure and required actions will be generated by the reliability
characteristics. A system model is developed and infor- study. It is also a help, of course, in defining the data
mation based on past performance of similar equipment requirements.
is then employed in the model to predict the probability
of failure in the critical system failure modes. Typically
the assessment will involve four main stages, namely :
2.3 Hazard and operability studies
System definition
Hazard and operability studies Understanding how the system functions in its different
Reliability modelling operational modes is obviously important. In very few
Sensitivity analysis cases is this information available at the start of a
reliability assessment. Where different disciplines are
The M S was received on 19 January 1993 and was accepted for publication on
involved quite different requirements for system oper-
20 January 1993. ation can be revealed. Many problems are in fact due to
E00293 0 IMechE 1993 G954-4089/93 $3.00 + .05 Proc Instn Mech Engrs Vol 207

Downloaded from at PENNSYLVANIA STATE UNIV on September 18, 2016


inconsistencies at interfaces-interfaces between disci- randomly distributed with time and (c) wear-out when
plines as well as between sub-systems. A simplified the equipment is no longer meeting its design specifi-
description of the plant and the operational require- cation and needs replacement.
ments in the different operating modes should always be For many system reliability assessments the initial
prepared at the start of the assessment and agreed with assumption is that all equipment operates in the useful
design, maintenance and operations engineers. life phase when the failure rate is virtually constant.
The reliability engineer will generally develop a sim- Consideration may be given at a later stage to the
plified functional block diagram to support the system effects of different (say, time-dependent) failure charac-
description. In many process plant design studies a teristics for the sub-systems identified as critical to plant
hazard and operability (HAZOP) study will also be operation.
carried out. Hazard and operability studies are well For equipment operating in the useful life phase the
described in the Chemical Industries Association guide probability of success (reliability) over a period of time t
(2). The HAZOP meetings will be chaired by a senior can be represented by the simple expression:
manager and involve engineers and scientists from rele-
R(t) = e-[
vant disciplines. The meetings will consider the conse-
quences of deviation from the design intention for each where e is the base of natural logarithms, ;1 is the
major equipment and sub-system. HAZOP studies are equipment failure rate and t is the time interval. Its
generally based on functional block diagrams and the complement, the probability of failure in time t, is
piping and instrumentation drawings. Guide words are
employed to stimulate the identification of possible F(t) = 1 - e-
failure modes and the consequences of deviations from These expressions refer to systems that are deemed
normal, steady state operation. Typically these guide unrepairable after failure. For repairable systems relia-
words are: bility is defined by the equipments steady state avail-
none no flow, no pressure, no cooling, etc. ability, where
more of more flow, temperature, pressure, etc. mean up-time - MTTF
less of reduced flow, temperature, pressure, A=
total time MTTF + MTTR
part of only some of the intentions are Here MTTF is the mean time to failure (= l/A) and
achieved, say component A is missing MTTR is the mean time to repair. Its complement,
from the process stream, etc. unavailability, is
other than say, impurities present, ingress of air,
water, acids, etc. mean down-time - MTTR
U =
total time MTTF + MTTR
The step-by-step analysis of the various sub-sections
of the plant by the HAZOP study provides a valuable The system reliability model is developed from the
insight into the potentially hazardous areas and oper- plant functional block diagram via system and sub-
ations. From the standpoint of the reliability engineer it system reliability block diagrams, failure or success
also helps to identify the important failure modes and logic diagrams or fault trees. These techniques involve
the sub-systems and equipment that need particular the logical combination of equipment failure probabil-
attention in the reliability assessment. ities (or unavailabilities) to generate sub-system and
system failure probabilities. The methods are fully
described with examples, in the IMechE Process
2.4 System reliability models Industries Divisions guide to reliability assessment (1).
Reliability can be defined as: the probability that an
item (system, equipment or component) will successfully
perform its required function for a specified period of
time under specified conditions. Clearly the probability 2.5 Sensitivity analysis
of success will change with time. For many equipments, The uncertainties associated with quantitative reliability
however, reliability behaviour can be represented by the assessments are due to two basic causes:
well-known reliability bath-tub curve shown in Fig. 1.
The three phases of the bath tub represent (a) burn-in (a) uncertainties in the reliability model,
when sub-standard components are weeded out, (b) (b) uncertainties in the data.
useful life when failures are relatively infrequent and
Uncertainties in the model reflect the assumptions
implicit in the system definition. Any representation of a
system will be a simplification so there will always be
some incompleteness in the reliability model. In most
cases, however, the objective of the reliability study will
be to compare different design alternatives; in these
situations the modelling uncertainties are likely to
cancel out.
I I I The effect of uncertainties in the data can be explored

I t by techniques such as sensitivity analysis. Sensitivity

analysis involves changing the value of the failure rate
Fig. 1 Reliability bath-tub curve (or unavailability) for each equipment in turn by a small
Part E: Journal of Process Mechanical Engineering Q IMechE 1993

Downloaded from at PENNSYLVANIA STATE UNIV on September 18, 2016


amount and then recalculating the system failure prob- failure, say an unrevealed failure of a control valve
ability. The effect of changes in equipment failure char- which inhibits operation of the plant fire water system.
acteristics can then be compared with the contribution One well-known data source gives an overall failure
from other equipments. rate of 30 failures/million hours for control valves.
The option to run a sensitivity analysis is often Others may give the proportion of failures in the differ-
included in the computer programs available to evalu- ent failure modes as
ate reliability models. In addition to routines for the
logical combination of failure events and calculations of Leakage 50%
system reliability the programs usually provide options Actuator 30%
to calculate the effect of changes in failure rates, Seize 10%
unavailabilities, repair times and test intervals for each Others 10%
equipment. The sensitivity for each failure event i will Not all of these failure modes will inhibit system oper-
be ation; for example leakage from the valve (unless gross)
is unlikely to materially affect the systems ability to
s.= % change in system failure probability provide fire water on demand. It is, therefore, necessary
% change in value of event i to adjust the control valve failure rate to reflect only
When the sensitivities have all been calculated they those failures that could contribute to a potentially
are listed in order of importance. It is then possible to dangerous failure of the fire water system. It may be
identify those equipments that have the most effect on assumed that actuator failures (30 per cent), seizure (10
the system reliability calculations. Where these effects per cent) and half the other failures (5 per cent) could
are significant it is worth while reviewing the data fail the fire water system so the control valve failure
sources employed for estimating specific equipment rate used in calculating the probability of the fire water
failure rates. At this stage generic data may not be ade- system failure on demand would be 13.5 failures/million
quate and it is then necessary to obtain a more repre- h (that is 45 per cent of 30 failures/million h). In all
sentative estimate by collecting and analysing failure reliability assessments it is important that the calculated
experience from similar applications or by techniques system failure probability is based only on the
such as failure mode and effect analysis. equipment failure modes that could provoke a system
failure. In most cases this failure rate is significantly less
than the overall failure rate of the equipment.
The data requirements for reliability studies are
shown in Fig. 2, taken from reference (l), which also
To calculate the reliability or availability of a system contains a more detailed discussion of the various
requires information on equipment failure rates and failure and repair statistics required in reliability assess-
repair times. It is important that these calculations are ments.
based on the equipment failure and repair modes that
have a direct impact on the system capability to meet a
specific functional requirement. These failure and repair
modes are seldom clearly defined in the reliability data
available. For example, interest may focus on failures Published sources of failure and repair statistics are
that could generate a potentially dangerous system quite limited, particularly for mechanical equipment. A

System Engineering Design/ Process

description Mean active Mean
disciplines construction parameters repair times waiting times
data data

Failure rates

Fig. 2 D a t a requirements for reliability assessment

@ IMechE 1993 Proc Instn Mech Engrs Vol 207

Downloaded from at PENNSYLVANIA STATE UNIV on September 18, 2016


Table 1 Principal reliability data sources

Data source Title Publisher and date
1. CCPS Guidelines for process equipment reliability American Institute of Chemical Engineers, 1990
2. Davenport and Warwick A further survey of pressure vessels in the AEA Technology-Safety and Reliability Directorate, 1991
UK 1983-1988
3. DEFSTAN 0041, Part 3 MOD practices and procedures for reliability Ministry of Defence, 1983
and maintainability, Part 3, Reliability
4. R. F. de la Mare Pipeline reliability; report 80-0572 Det Norske Veritas/Bradford University, 1980
5 . Dexter and Perkins Component failure rate data with potential EI Du Pont de Nemours and Company, USA, 1982
applicability to a nuclear fuel reprocessing
plant, report DP-1633
6. EIREDA European industry reliability data handbook, EUROSTAT, Paris, 1991
Vol. 1, Electrical power plants
7. EN1 Data Book EN1 reliability data bank-component Ente Nazionale Indrocarburi (ENI), Milan, 1982
reliability handbook
8. Green and Bourne Reliability technology Wiley Interscience, London, 1972
9. IAEA TECDOC 478 Component reliability data for use in International Atomic Energy Agency, Vienna, 1988
probabilistic safety assessment
10. IEEE Std 500-1984 IEEE guide to the collection and presentation Institution of Electrical and Electronic Engineers, New
of electrical, electronic sensing component York, 1983
and mechanical equipment reliability data
for nuclear power generating stations
11. F. P. Lees Loss prevention in the process industries Butterworth, London 1980
12. MIL-HDBK 217E Military handbook-reliability prediction of US Department of Defense, 1986
electronic equipment, Issue E
13. NPRD-3 Non-electronic component reliability data Reliability Analysis Centre, RADC, New York, 1985
14. OREDA 84 Ofshore relibility data (OREDA) handbook OREDA, Hovik, Norway, 1984
15. OREDA 92 Offshore reliability data, 2nd edition DnV Technica, Norway, 1992
16. RKSjSKI 85-25 Reliability data book for components in Swedish RKS-Nuclear Safety Board of the Swedish Utilities and
nuclear power plants SKI-Swedish Nuclear Power Inspectorate, 1987
17. H. A. Rothbart Mechanical design and systems handbook McGraw-Hill, 1964
18. D. J. Smith Reliability and maintainability in perspective Macmillan, London, 1985
19. Smith and Warwick A survey of defects in pressure vessels in the AEA Technology-Safety and Reliability Directorate, 1981
UK (1962-1978) and its relevance to primary
circuits, report SRD R203
20. WASH 1400 Reactor safety study. An assessment of US Atomic Energy Commission, 1974
accident risks in U S commercial nuclear
power plants, Appendix 111, Failure data

sub-group of the IMechE Mechanical Reliability Com- 5 GENERIC RELIABILITY DATA ANALYSIS
mittee has therefore carried out a number of studies to
identify and evaluate the main sources of mechanical For new plant designs there is clearly no experience of
reliability data currently available. This information will actual reliability performance that can be used in the
be published later this year as a series of data sheets in assessment. In these circumstances it is necessary to
the second edition of the Process Industries Divisions employ generic data from source documents such as
guide to reliability assessment (1). The data sheets were those listed in Table 1. These data will also need adjust-
prepared by members of the Data Acquisition Working ment for the differences in the operating duty and
Party who are all practising reliability engineers. The environmental conditions of the proposed design and
principal sources identified and evaluated by the those reflected by the generic data. Unfortunately the
Working Party are shown in Table 1. The majority are operational and environmental conditions are seldom
in the public domain. Although the Working Party has identifiable from the source data. Assumptions must
concentrated on reviewing sources of mechanical rather therefore be made and engineering judgement applied
than electronic data, many of the sources include both. when compiling a reliability database for any particular
The cross reference table (Table 2) indicates the type of safety or availability assessment.
equipment covered in the data source documents. Table The sources of data will also be of varying quality. In
3 shows typical components within the equipment cate- the best cases the data will be comprehensive with infor-
gories shown in Table 2. An example data sheet is mation on the engineering and functional characteristics
shown in Fig. 3. of the equipment to complement the estimates of failure
In preparing the data sheets the Working Party has rate in the principal failure modes. This engineering and
not offered any judgement as to the quality, applicabil- functional information could include definitions of the
ity or validity of the data contained within them. The boundaries of the equipment, the environment and
sole purpose has been to provide guidance on a number operational duty, the number of failures observed in
of useful sources and their content for engineers with each failure mode, the size of the equipment sample
a need for numerical mechanical reliability data. population and the length of the surveillance period. In
However, it must be said that it is extremely unlikely other cases, however, the information will be very
that better data sources exist in the published literature restricted, possibly confined to an overall failure rate
at the present time. estimate for a general class of equipment. Thus, when
Part E: Journal of Process Mechanical Engineering @ IMechE 1993

Downloaded from at PENNSYLVANIA STATE UNIV on September 18, 2016


Table 2 Cross reference between principle data sources-equipment type

Reference Application* Rotating Static Instrument Safety Process Electrical
2. Davenport and Warwick NI - Y - - - -
3. DEFSTAN 0041, Part 3 M Y Y Y Y Y Y
4. de la Mare 0 - Y - - - -
5. Dexter and Perkins, DP-1633 NI Y Y Y Y Y Y
7. EN1 I Y Y Y Y Y Y
8. Green and Bourne NI Y Y Y Y Y Y
10. IEEE Std 500 I Y - Y Y - Y
11. Lees NI Y Y Y Y Y Y
12. MIL-HDBK 217E M Y Y - Y - Y
13. NPRD-3 NI Y Y Y Y - Y
14. OREDA 84 0 Y Y Y Y Y Y
15. OREDA 92 0 Y Y Y Y Y Y
16. RKS 85-25 NI Y Y Y Y - Y
17. Rothbart I Y Y Y Y - Y
18. Smith NI Y Y Y Y Y Y
19. Smith and Warwick NI - Y - - - -
20. WASH 1400 N Y Y Y Y - Y
* 0 = offshore, N = nuclear, I = industrial, M = military.
Note: Table 2 should be used in conjunction with Table 1.

Table 3 Typical components within equipment categories assembling a reliability database for a specific study the
Equipment category Typical components objective should always be to identify as many represen-
tative data sources as possible and to employ the com-
1. Rotating equipment Pumps, compressors, turbines, motors
2. Static equipment Pipes, pipelines, flowlines, valves, vessels bined information to derive best-estimates of the
3. Instrumentation Sensors (temperature, pressure, level, equipment failure rate and its principal failure modes
etc.) and controllers before making adjustments for specific applications.
4. Safety Fire pumps, safety valves, fire/gas/smoke For electronic components where the sample data are
5. Process Pumps, compressors, valves, vessels,
compiled from tests on large populations of com-
piping ponents under controlled conditions, it is reasonable to
6. Electrical Cables, motors, circuit boards, lamps treat the data from a number of different sources as a
Note: it should be noted that these categories are not intended to be mutually
single homogeneous sample and calculate a mean
exclusive. For example, pumps can be categorized as both rotating and process. failure rate. For mechanical equipment the situation is

Data source: CCPS

Full name: Guidelines for process equipment reliability, with data tables prepared by the
Equipment Reliability Subcommittee of the Centre for Chemical Process Safety
and Science Applications International Corporation
Published by: American Institute of Chemical Engineers
Type of source: Book
Date of publication: 1990
Type of components: A wide range of chemical process equipment
Number of types of components: 76 process equipment types
Type of information: Failure rates based on both calender and operating time; failures on demand are also
included where appropriate
Source of data: Data sources include plant specific data and generic data. Plant specific data reflects
plants process, environment and maintenance practices. Generic data was collated
from a variety of sources. All major published sources of available generic
equipment reliability and failure rate data were used, including reliability studies,
published research works, reliability data banks, government reports containing
information gathered from chemical process, nuclear, offshore oil and fossil fuel
industries around the world.
Type of data: The data are characterized as equipment failures per lo6 operating hours for
time-related failure rates and failures per lo3 demands for demand-related failure
rates. Rates are given for common chemical process equipment. Equipment used
for transport of chemicals are not covered. The cause of equipment failures, the
means to improve reliability and the most reliable equipment are not addressed.

Fig. 3 Example 1 of 20 source data sheets

0 IMechE 1993 Proc lnstn Mech Engrs Vol 207

Downloaded from at PENNSYLVANIA STATE UNIV on September 18, 2016


quite different since the majority of sources will reflect Table 4 Environmental stress factors ( k , )
experience from plant operating in significantly different General environmental conditions k.
conditions. The sample sizes will frequently be small
and there may be substantial differences in the design of Ideal, static conditions 0.1
Vibration-free, controlled environment 0.5
the equipment, the size, duty, etc., and in the operating Average industrial condition 1.o
and environmental conditions. Such divergent source Chemical plant 1.5
data clearly cannot be combined as a single homoge- Offshore platforms/ships 2.0
neous sample. It would also exclude failure rate esti- Road transport 3.0
Rail transport 4.0
mates from sources where details of the sample Air transport (take-ofl) 1LO.0
populations, numbers of failures, surveillance period,
etc., are not recorded in the data source document.
Because the sources of mechanical reliability data are Table 5 Rating stress factors ( k 2 )
relatively few, this could seriously diminish the generic
Component nominal rating
data available to the reliability assessor.
Combining reliability data from a number of different Yo k,
sources has been studied by several different organiz- 140 4.0
ations, for example in the Offshore Reliability Data 120 2.0
100 1.o
(OREDA) Project (3). Moss, in a study of process 80 0.6
control equipment reliability, also found that sample 60 0.3
failure rates for many equipment classes in a variety of 40 0.2
applications were generally log-normally distributed (4). 20 0.1
For most types of equipment over 60 per cent of sample Note: for equipment designed aganst a specific
failure rates were within a factor of 2 and over 90 per design code (for example vessels) rating = 100%;
for other equipment (for example pumps, valves,
cent within a factor of 4 of the median failure rate. etc.) rating = 80% or less may be appropriate.
As a first approximation, therefore, it may be
assumed that a best-estimate of the mean failure rate for
average industrial conditions will be the median of a set From the foregoing a more robust estimate of the
of sample failure rates since 50 per cent are less than failure rate for the fire water system control valve given
and 50 per cent more than this value. The range of in Section 3 can be derived based on the additional
sample failure rates also defines the uncertainty associ- information given in Table 6.
ated with the estimate due to the effect of different oper-
ating and environmental conditions and the functional Table 6 Application details
and engineering characteristics of the equipment. This Equipment Control valve, normally closed
can be useful information to support the sensitivity Function Open on demand
analysis. TYPe Pneumatically actuated, gate valve, 2 in diameter
The base failure rate will, however, need adjusting Operating duty River water, 100 Ib/in2, ambient temperature
Environment Chemical plant
for the specific application being assessed and the Failure mode Fail-closed (FC)
equipment failure modes which can cause failure of the
system. Models for adjusting equipment failure rates are
generally of the form Consider first the data available from the source
documents listed in Table 2. It can be seen that 13 of
AXA = nn

i= 1
ki dA) the data sources feature static equipment in an indus-
trial environment and closer inspection of the data
sheets shows that all of these sources contain some
where A, is the predicted failure rate for equipment X information on valves. From the data tabulated in the
in failure mode A, 1, is the base failure rate for source documents five sources specifically quote failure
equipment types similar to X, ki is the stress factor for rates for control valves-some of the data sources also
stress i and p(A) is the proportion of failures in failure give details of the valve size, operating medium, valve
mode A. Here ll is the multiplication operator. The type, method of actuation and the principal failure
importance of applying a factor p(A) for the proportion modes. However, the extent of the information from the
of equipment failures that can cause failure of the different sources varies widely, in some cases (source 8)
system cannot be overemphasized. Without this factor just an overall failure rate is given, in others (source 9)
the estimate of I,,is likely to be unduly pessimistic. the data includes details of the valve size, type, oper-
For global data such as those available from the data ating environment, sample size, number of failures, etc.
sources featured here a simplified model is usually After careful review of the source documents five
employed samples were selected for deriving a best-estimate of the
base failure rate for this application. The overall failure
rate estimates from these five sources ordered by magni-
where k , is the stress factor for the environment and k , tude are 2.4, 10, 18.7, 30 and 45 failures/million h. The
is the stress factor for equipment duty. Tables 4 and 5 median failure rate of 18.7 failures/million h is therefore
show recommended stress factors for environment and a best-estimate of 1, for average industrial conditions.
duty respectively. The proportion of failures in the rele- The stress factors for environment and duty can be
vant failure mode p(A) may be derived from the source obtained from Tables 4 and 5. This chemical plant can
data or alternatively may be estimated from engineering be assumed to be typical of chemical plants generally;
experience. hence the environmental stress factor k , = 1.5 for this
Part E: Journal of Process Mechanical Engineering @ IMechE 1993

Downloaded from at PENNSYLVANIA STATE UNIV on September 18, 2016


Table 7 Control valve-DroDortion of failures in fail-closed mode

Assumed failure modes Percentage Effect of Estimated proportion
(engineering judgement) failure in fail-closed mode
Leakage 50% No effect 0%
Actuator failures 30% Will not open 30%
Seize 10% Will not open 10%
Others 10% Assume half in fail-closed mode 5 Yo
Total 45%

application. The duty specified for this control valve tive failure and repair statistics. Those listed here are
does not appear to be particularly onerous and engin- considered to be the most comprehensive sources avail-
eering judgement suggests a duty stress factor k, = 0.6; able from the public domain.
this assumes that the operating conditions are signifi- Consideration is currently being given by the Mecha-
cantly lower than the valve design specification as far as nical Reliability Committee to the preparation of a
operating pressure and temperature are concerned. handbook of reliability data for the process industries.
The proportion of failures in the different failure The handbook will be based on the data sources listed
modes given previously can be accepted as representa- in Table 1, in-house reliability data from companies in
tive after comparison with the information from the five the process industries, existing reliability studies of
source documents selected for use with this study. It is process equipment and expert judgement. These data
now necessary to estimate the proportion that could will be combined to generate best-estimates of failure
contribute to system failure-in this case control valve rates, failure modes, failure on demand probabilities
fails closed. Table 7 shows how this can be estimated. and repair times for a wide range of mechanical
Thus for this fire water system control valve best- equipment. The authors would welcome contributions
estimates are as follows: from engineers with access to alternative sources of
Base failure rate (failures/million h) & = 18.7 process equipment reliability information.
Stress factor k , = 1.5
Duty stress factor k, = 0.6
Proportion in fail-closed mode p(FC) = 0.45
Hence The data source sheets discussed in this paper were pre-
pared by the Mechanical Reliability Committee Data
= I b k l k , X p(FC) Acquisition Working Party with the following member-
= 18.7 x 1.5 x 0.6 x 0.45 = 7.6 failures/million h ship :
Dr J. E. Strutt, Chairman, Cranfield Institute of Tech-
This is significantly lower than the estimate obtained in nology
Section 3 based on only one data source. Failure rate E. A. Boxall, Nuclear Electric
estimates based on limited information generally tend D. Thompson, BP International Limited
to be pessimistic. The benefit of using several sources Dr D. W. Heckle, ICI Limited
should always produce more realistic figures and an P. Stead, BEQE Limited
indication of the uncertainty in the estimate. However, L. Dunbar, RM Consultants Limited
some of the source documents discussed here are quite B. J. Hanks, British Gas Corporation
specific to one industry and are based on extensive data
collection and analysis exercises. One such example is
source 15, the Offshore reliability data handbook. In
these cases the data source may be sufficient in itself, REFERENCES
although other data sources could be useful for com- 1 Davidson, J. and Hunsley, C. The reliability of mechanical systems,
ponents not covered in the handbook and for compari- 2nd edition, IMechE Guides for the Process Industries, 1993
son. (Mechanical Engineering Publications, London).
2 A guide to hazard and operability studies, 1979 (Chemical Industries
6 CONCLUSIONS 3 Oflshore reliability data (OREDA) handbook (OREDA 84), 1984
(OREDA Participants, Norway).
Quantitative reliability assessment of mechanical 4 MOSS, T. R. The reliability characteristics of process control
systems is frequently inhibited by the lack of representa- equipment. MPhil thesis, University of Bradford, 1981.

@ IMechE 1993 Proc Instn Mech Engrs Vol 207

Downloaded from at PENNSYLVANIA STATE UNIV on September 18, 2016