OREDA-2009 1 Volume 1— Topside Equipment INTRODUCTION THE OREDA PROJECT The Offshore Reliability Data (OREDA) project was established in 1981 in co-operation with the Norwegian Petroleum Directorate’, The initial objective of OREDA was to collect reliability data for safety equipment. The current organisation, as a co-operating group of several oil companies, was established in 1983, and at the same time the scope of OREDA was extended to cover reliability data from a wide range of equipment used in oil and gas exploration and production (E&P). Offshore topside and subsea equipment are primarily covered, but some onshore E&P equipment is also included and data collection is currently being extended for this industry. The main objective of the OREDA project is to contribute to an improved safety and cost- effectiveness in design and operation of oil and gas E&P facilities; through collection and analysis of maintenance and operational data, establishment of a high quality reliability database, and exchange of reliability, availability, maintenance and safety (RAMS) technology among the participating companies. PROJECT PHASES Phase I (1983 - 1985) The purpose of phase J was to collect and compile data from offshore drilling and pro- duction operations. The data were published in the OREDA-84 handbook. An objective of the handbook was to demonstrate the ability of the eight participating oil companies to co- operate on this issue and create a forum for a common co-operative process in this field. Data was collected on a wide area of equipment (large population) but not with as much detailed information as in later phases. Data from this phase is not included in the OREDA computerised database. Phase II (1987 - 1990) The scope was adjusted to only collect data on production critical equipment, to improve the quality of the data, and to store the data in a PC database format. A tailor-made PC program (called the OREDA software) was developed to aid the collection and analysis of the data. The data were published in the OREDA-92 handbook. This handbook also contains the data collected in phase I. Phase IIT (1990 - 1992) The number of equipment categories was increased, and more data on maintenance programs were collected. The data quality was improved by means of the comprehensive “Guidelines for Data Collection” and through quality control. The OREDA software was modified into a more general-purpose data collection tool, and its user interface was improved. The data collected in this phase are published in the OREDA-97 handbook. ? Has since then changed organization and name to Petroleum Safety Authority Norway @OREDAVolume 1 - Topside Equipment 12 OREDA-2009 Phase IV (1993 - 1996) New software for data collection and analysis was developed, plus specific software and procedures for automatic data import and conversion. Data were collected mainly for the same equipment as in phase III, and the data collection was - to a greater extent - carried out by the companies themselves. Data on planned maintenance were included. The data collected in this phase are published in the OREDA-02 handbook. Phase V (1997 - 2000) Some new equipment classes were included and more focus was given on collecting subsea data. As a parallel activity, the ISO standard 14 224: “Petroleum and natural gas industries - Collection and exchange of reliability and maintenance data for equipment” was developed and issued in July 1999. The data collected in this phase are published in the OREDA-02 handbook. Phase VI (2000 - 2001) Data collection on subsea equipment and new equipment classes were prioritised. A forum for co-operation between major subsea manufacturers was formed. Phase VII (2002 - 2003) Priority was given on safety and subsca equipment. A revision of ISO 14224 was started with important contribution from members of the OREDA project. The final version was issued in December 2006. ‘The OREDA project has since continued and is currently (2009) in its tenth phase. Up-to-date information on the OREDA project is available on the Internet address: http://www.oreda.com PARTICIPANTS During phase VI and VII there have been several changes among OREDA participants. Table 1 summarises the companies that have contributed with data in these phases. © OREDAOREDA-2009 13 Volume 1 - Topside Equipment Table 1 - Companies that have Contributed with Data in Phase VI and VII Companies Phase VI Phase Vil. | Comments Eni/Agip 1 Vv BP a v Chevron qv Merged with Texaco and left OREDA in phase VI ExxonMobil x y Exxon and Mobil merged in phase VI! Norsk Hydro v Vv Merged with Statoil in 2007 ConocoPhillips Petroleum Vv v Phillips and Conoco merged in phase Vil Company Norway StatoilHydro V Vv StatoilHydro changed name to Statoil Nov. 1" 2009 Shell v Vv Texaco V Merged with Chevron and left OREDA in phase VI Total qv v Merged with Elf and Fina in phase VI ORGANISATION OREDA is managed by a Steering Committee with one member and one deputy member from each of the participating oil companies. The Steering Committee elects one of its members as chairman and appoints a Project Manager. The Project Manager co-ordinates the activities approved by the Steering Committee. and also performs data quality assurance. Det Norske Veritas (DNV) served as Project Manager during phases I and I] and SINTEF during phases I] — IX, DNV is the current Project Manager of phase X which was started in 2009. The preparation of the OREDA handbooks has been carried out as separate projects in agreement and consultation with the OREDA Steering Committee. The current version has been prepared by SINTEF and is marketed by DNV. Equipment CATEGORIES COVERED INOREDA HANDBOOK Table 2 shows the equipment population that has been included in the 2009 handbooks. As focus in OREDA phases VI and VII has been on collecting subsea data, there are some topside equipment classes for which new data have not been collected. To obtain a reasonable population for presenting reliability data for topside equipment in the 2009 edition, some data issued in the previous 2002 handbook have been included (see Table 2). For topside equipment, where no reliability data have been collected in phases VI and VII, the 2009 edition corresponds to the 2002 edition; however, reliability data is presented at a more coarse taxonomy level than in the 2002 edition. Most of the equipment derives from offshore facilities, but a few equipment units from onshore E&P have also been included. © OREDAVolume 1 - Topside Equipment 14 OREDA-2009 Table 2 — Equipment Population in the 2009 OREDA Handbooks Topside Handbook (Volume 4) Data from 2002 edition | New datain | Total 2009 included in 2009 edition | 2009 edition | edition. SYSTEM Tae Phaseiv | Phase | Phasas vi« No. ofunits | No. ofunits | No. ofunits | No. of units 1, Machinery 1.1 Compressors 7 56 131 1.2 Gas Turbines 56 32 88 1.3 Pumps 160 52 212 1,4 Combustion Engines 75 23 98 1.5 Turboexpanders 8 2 10 2. Electric 2.1 Electric Generators 26 6 32 Equipment _|2.2 Electric Motors 128 45 143 3. Mechanical 3.1 Heat Exchangers 17 4 21 Equipment 3.2 Vessels 148 50 198 3.3 Heaters and Boilers ail 1 42 4, Other Topside | 4.1 Fire & Gas Detectors 779 139 918 Equipment —_| 4.2 Process Sensors 6 69 4.3 Control Logic Units* 10 10 4.4 -4.5 Valves* 334 576 907 ‘SUM TOPSIDE 290 1732 827 2849 ‘Subsea Handbook (Volume 2) 5, Subsea 5.1 Control Systems 7 105 422 5.2 Flowlines 59 251 310. 5.3 Manifolds 29 7 106 5.4 Pipelines (SSIV) 85 4 89 5.5 Risers 42 104 146 5.6 Running Tools 5 5 5.7 Templates 45 15 5.8 Wellheads & X-mas Trees 83 479 262 SUM SUBSEA 320 735, 41055, Note 1: Data from phase V issued in the 2002 handbook have been included in the population of this handbook. For equipment where no new data have been collected since the 2002 issue, data from phase IV and phase V previously issued in the 2002 issue have been included. Since the 2002 issue some changes in the equipment population for phase IV and V data has been made due to additional new data, removal of outdated equipment etc. Note 2: The equipment population for phases lV and V shown in above table only represent the population that have been included in the 2009 handbook issue, not the total population in the OREDA database. * Contro! logic units were not included in the 2002 edition. * valves are described in two different ways, i.e. by application code (4,4) and by taxonomy code (4.5). It should be noted that the corresponding taxonomy number for valves in the 2002 edition were 4.4 and 4.3, respectively. ®OREDAOREDA-2009 15 Volume 1— Topside Equipment Score of OREDA HANDBOOK For each sopside equipment unit, the following information is presented: * A drawing illustrating the boundary of the equipment unit and specification of subunits and maintainable items that are part of the various subunits. e A listing of all failure modes, classified as Critical, Degraded, Incipient or Unknown. © The aggregated observed time in service for the equipment unit, classified as Calendar time, Operational time and Number of demands. © The observed number of failures for each failure mode. * An estimate of the constant failure rate for each failure mode with associated uncertainty intervals. © Mean and maximum values of the repair time, i.e. the number of man-hours required to repair the failure and restore the function. © Mean and maximum values of the active repair time, i.e. the elapsed time in hours to repair the failure and restore the function (time when actual repair work was being done). © Supportive information, e.g. equipment population and number of installations. * A cross-tabulation of: a) Failure mechanism/cause versus Failure mode b) Failure mechanism/cause versus Severity LIMITATIONS Information released from each participating company has been kept confidential by rendering it anonymous. Only generic data are published. Failure and maintenance data are in most cases gathered from two or more installations, and consequently the figures in the handbook reflect a weighted average of the experience, The OREDA project is so far restricted to failure data collected on hardware equipment. Failures initiated by humans are, implicitly, included in the failure rate estimates. All data shown in the -09 edition are extracted from the OREDA database before the OREDA taxonomies were updated to comply with ISO 14224 (phases > VIII). Details and limitations of the methods used are described in the section “ESTIMATION PROCEDURES”, p. 23. ®OREDAVolume 1 - Topside Equipment 16 OREDA-2009 (This page intentionally left blank) @OREDAOREDA-2009 17 Volume 1 — Topside Equipment OREDA TOPSIDE DATA STRUCTURE The structure of the OREDA topside database is described to better understand how the reliability data presented in this handbook are extracted and compiled. Main DATA CATEGORIES For each equipment category the database is split into three separate database files: an Inveniory part, a Failure part, and a Maintenance part. The Inventory part contains a description of each Equipment unit for which data have been collected. This description contains technical data (e.g. capacity, size) as well as some operating and environmental data (e.g. operating mode, vibrations). The inventory description for each equipment unit is stored in an Jnventory record in the database. The Failure part contains information about the failures recorded for the equipment unit (inventory) during the period of surveillance; one record for each failure event. The failure events are always related to one equipment unit. In OREDA a failure event is defined as a physical failure of equipment. This implies that all events where a work order is issued, and some maintenance action carried out, would be considered as failure (see the definitions on page 41). The Maintenance part contains information about corrective and scheduled preventive maintenance program for each equipment unit (c.g. maintenance action, interval, man- hours). Data on corrective maintenance is related to its preceding failure, while data on preventive maintenance is related to the equipment unit. The information recorded in the OREDA database is partly based on numeric data, partly on codes selected from a predefined menu, and partly on free text (see Table 3). @OREDAVolume 1 - Topside Equipment 18 OREDA-2009 ‘Table 3 — Failure and Maintenance Data Collected in OREDA Database | FAILURE MAINTENANCE Description Format Description Format Identification | Failure record Numeric | Maintenance record number Numeric data number Equipment unit name. | Text Corresponding failure record no Numeric Work order number | Numeric | Equipment unit name Text Status Code tist__| Status Code list Event data | Failure detection date | Date Maintenance date Date Failure mode Code list. | Maintenance category Code list Severity class Code list. | Subunit maintained Code list Failure mechanism | Code list | Items maintained Code list Failure cause Code list. | Maintenance action Code list ‘Subunit failed Code list | Maintenance interval (periodic Numeric maintenance only) Maintainable item(s) | Code list Maintenance manhours, per discipline and | Numeric failed total Failure detection Code list. | Active maintenance time Numeric method Failure consequence | Code list_| Downtime Numeric Additional | Specific remarks Code list. | Specific remarks code Code list information | code Additional free text | Text Additional free text info Text info SYSTEM HIERARCHY The topside system hierarchy is illustrated in Figure 1, p. 20, Equipment Category: Items are grouped into Equipment Class according to main function of the item (e.g. pumps, compressors, valves ete.). Items of an Equipment Class are further classified according to design characteristics and type of service. Table 4 gives an example for the equipment class Pumps and its corresponding Design Class and Service (system) (e.g. pump, centrifugal, combined function) with taxonomy no. Equipment Unit; Each individual item within an Equipment Class (c.g, a pump) Subunit: An item required for the Equipment Unit to perform its main function (e.g. power transmission, lubrication, control and monitoring). The subunits may be redundant, e.g. two independent start units. Maintainable Item (MI): A subset of each Suhunit that will typically consist of the lowest level items that are due for preventive maintenance. © OREDAOREDA-2009 19 Volume 1 — Topside Equipment ‘Table 4 — Equipment Categories (Example) EQUIPMENT CLASS DESIGN CLASS SYSTEM Description Taxonomy, Description oe Description chore no no Pumps 13 Centrifugal 1.3.1 | Combined function 13.44 Condensate processing | 1.3.1.2 Cooling systems 1343 Crude oil handling 1344 Flare, vent & blow-down | 1.3.1.5 Gas processing 43.46 Gas treatment 1.3.1.7 Heating medium 1348 il export 1.3.1.9 Oil processing 4.3.4.10 Oily water treatment 1.3.4.41 Seawater lift 4.3.1.12 Water fre fighting 4.34.43 Water injection 43.4.14 Reciprocating 4.3.2 | Chemical injection 4324 Gas processing 4.322 Gas treatment 1323 Main power 4.324 Water fire fighting 13.25 Rotary 1.3.3 | Oily water treatment 4.3.3.4 (Some equipment classes (e.g. Electric generators) are further classified according to the size of the equipment.) ©OREDAVolume 1 - Topside Equipment 20 OREDA-2009 Maintainable item Lt p> Subunit 2.7 — >| Equipment unit 2 |) ben Lf Subunit 1.2 Y Maintainable item System _— +. | Lp | Subunit /.m ’ ial Subunit 7.7 LT Equipment wait [- Subunit 7.2 _-»| Maintainable item nm. Maintainable item % nm. ‘Subunit 2.0 7 Maintainable item nt. Figure 1 - Equipment hierarchy Some subunits (e.g. lubrication system) may be relevant for several equipment classes. In these cases the subunits are given the same name and the same set of Ms. This is done in order to standardise the subunits/MIs as much as possible, although some of the Mis in these subunits may not apply for all equipment classes, EQuiPMENT BOUNDARIES For each Equipment Class, an Equipment Boundary has been defined to identify items that are part of the Eguipment Class and to show the interface between these items and their surroundings. Items within Equipment Boundary are considered to be essential for the function of the Eguipment Unit and/or they are typically part of the Equipment Unit as sold by the manufacturer. For example, the power transmission (e.g. gear) is included within the boundary for a pump, while the driver (e.g. electric motor) is not. The boundary is normally sufficiently determined by a boundary diagram as illustrated in Figure 2, p. 21. Further, a tabular description as shown in Table 5 lists those subunits and MIs that are included within the boundary. When establishing the boundaries, considerations were given to tag number principles used by the participating oil companies. Generally, equipment units correspond to the companies" main tag level, while subunits correspond to sub-tag levels. The following principles have been applied: @OREDAOREDA-2009 21 Volume 1 — Topside Equipment © Connected units are generally not considered to be part of the equipment unit, Failures that occur ina connection (e.g. leak) are included unless it is known specifically that it has occurred on the connected item outside the boundary. e When a driver and the driven unit use common subunits (e.g. lubrication), failures of the subunit are generally related to the driven unit. * Failures on drivers (e.g. gas turbine) and driven units (e.g. compressors) are recorded for each equipment class separately, When e.g. a failure rate for a combination of driver and driven units is needed (e.g. compressors driven by gas turbines) the combined values from those two equipment classes should be used. e Failures on instrumentation are only included if the instrumentation has specific control and/or monitoring function for the equipment unit and/or is locally mounted (e.g. sensors). Instrumentation of a more general use, such as supervisory system (SCADA) is not included. ae 4 i | power priverR | } ‘STARTING i | TRANS- ‘SYSTEM (Diesel El. TE] MISSION PION motor, ete.) | 1 | (Gearbox. etc.) | i i | } CENTRE LUBRICATION misc. i ano SYSTEM i MONITORING i { i | i Power Cooiant Boundary Figure 2 — Boundary definition for pumps (example) ©OREDAVolume 1 - Topside Equipment 22 OREDA-2009 Table 5 — Maintainable Items for Pumps (Example) EQUIPMENT CLASS PUMPS Subunit Power ‘ Control and Lubrication cA Pump unit he Miscellaneous transmission monitoring system Maintainable | Bearing Casing Actuating device —_| Cooler Coolingiheating ene Coupling to. Cylinder liner Control unit Filter system aan tam Diaphragm Intemal power Motor eocone gaperatyy Counting 2 impor supoly 5 Flange joints ek, Piping Monitoring Piping Pulsation damper variable drive | Piston Piping Pump Puryear Lubrication Radial bearing | Seals Reservoir (incl. ea Seals Sensors" heating system) hse Valves Seals Support Wiring Valves Thrust bearing Valves * Type of instrumentisensor, e.g. pressure, temperature, level etc. @OREDAOREDA-2009 23 Volume 1 — Topside Equipment ESTIMATION PROCEDURES The main purpose of the OREDA-2009 handbook is to present average failure rate estimates together with repair time estimates. This section presents a brief description of the statistical methods that are used_ FAILURE RATE ‘The failure rate function expresses how likely it is that an item that has survived up to time 1, will fail during the next unit of time. If the item is deteriorating, this likelihood will increase with the age ¢ A man who has reached the age of 95 years will obviously have a higher probability of dying during the next year than a 20 years old man, The failure rate function will therefore usually be a function of the time - or, the age of the item. To give a mathematical definition of the failure rate function, we start with the time to failure, 7, of the item, i.e. the time from the item is put into operation until the first failure occurs, It is generally impossible to predict the exact value of the time to failure, and 7 will therefore be a random variable with an associated distribution. The failure rate function, 2(1), may now be defined mathematically as: 1 20) = lim P< T <4 AT >0) This implies the approximation 2(f)-At = P(tt) The right hand side of this equation denotes “the probability that the item will fail in the time interval (7, f + Af), when the item is still functioning at time /” in other words: “the probability that an item that has reached the age / will fail in the next interval (f, ¢ + Ad).” The approximation is sufficiently accurate when Ar is the length of a very “short” time interval. The failure rate function is sometimes also called ‘hazard rate’ or ‘force of mortality’. The life of a technical item may generally be split into three different phases: the burn-in (or ‘early failure’) phase, the useful life phase and the wear-out phase. The failure rate function will usually have different shapes in the three phases. As illustrated in Figure 3, p. 24, the failure rate function may be decreasing in the burn-in phase, close to constant in the useful life phase and increasing in the wear-out phase. The curve in Figure 3 is called a “bath-tub” curve because of its characteristic shape, and is often claimed to be a realistic model for mechanical equipment. ©OREDAVolume 1 - Topside Equipment 24 OREDA-2009 If we assume that the failure rate function is constant during the useful life phase, provided it is properly maintained, this means that the item is not deteriorating during this phase. The deterioration will start when, or if, the item enters the wear-out phase. Failure rate finetion 2(7) Bum-in phase ‘Useful life phase Wear-out pliase Time ¢ Figure 3 — Bath-tub shape of the failure rate So-called burn-in problems may be caused by inherent quality problems in the item, or by installation problems. Inherent quality problems may sometimes be removed by carefull quality testing prior to installation. Installation problems have been disregarded in the OREDA data collection, notably for most /opside equipment. The burn-in phase is therefore not included in the OREDA database, and we may assume that the data collection is started with the useful life phase. Many of the items covered in OREDA are subject to some maintenance or replacement policy. The items will therefore often be replaced or refurbished before they reach the wear- out phase. The main part of the failure events in the OREDA database will therefore come from the useful life phase, where the failure rate is close to constant. All the failure rate estimates presented in this handbook are therefore based on the assumption that the failure rate function is constant and independent of time, in which case 2(t) = A, i.e. the failure rates are assumed to be exponential distributed with parameter 2. Note: e No statistical tests have been performed to verify the assumption of a constant failure rate. © Since data are assumed to come from the “bottom” part of the bath-tub curve, the failure rate estimates presented therefore represent some kind of minimum over the entire life cycle of the equipment. An important implication of the constant failure rate assumption is that an item is considered to be “as good as new” as long as it is functioning. All failures are purely chance failures and independent of the age of the item. @OREDAOREDA-2009 25 Volume 1 - Topside Equipment Based on the assumption of an estimated constant failure rate, the mean time to failure, MTTF, may be calculated as: MTTF These and related concepts are thoroughly discussed in e.g. Rausand and Hayland (2004). EsTIMATORS AND CONFIDENCE INTERVALS FOR A HOMOGENEOUS SAMPLE When we have failure data from identical items that have been operating under the same operational and environmental conditions, we have a so-called homogeneous sample. The only data we then need in order to estimate the failure rate A are the observed number of failures, n, and the aggregated time in service, 7. The maximum likelihood estimator of A is given by: i Number of failures Aggregated timein service 7 The aggregated time in service, t, may be measured either as calendar time or operational time, and both these are presented in the data tables in Part II Note that this approach is valid only in the following situations: e Failure times for a specified number of items, with the same failure rate A, are available. * Data (several failures) is available for one item for a period of time, and the failure rate Ais constant during this period. * A combination of the two above situations, i.c. there are several items where each item might have several failures. This is the typical situation for the OREDA data. In the data tables in Part II of the handbook, estimates are given for each failure mode. Confidence intervals for the failure rate The uncertainty of the estimate A may be presented as a 90% confidence interval. This is an interval (A;,2y). such that the “true value” of A fulfils: P(A; $ A< Ay) = 90% With 7 failures during an aggregated time in service 7, this 90% confidence interval is given by: l a ato) © OREDAVolume 1 - Topside Equipment 26 OREDA-2009 where zoos, and zoos, denote the upper 95% and 5% percentiles, respectively, of the x- distribution with v degrees of freedom, see Table 6, p. 33. Example Assume that n = 6 failures have been observed during an aggregated time in service T = 10000 hours. The failure rate estimate is then given by: A =n/t=6-10" failures per hour and a 90% confidence interval is given by: 1 1 1 1 4 a Fy Zon D— Zosriner) || SO 209512 So = (2.6-10°,11.8-10 The estimate and the confidence interval are illustrated in Figure 4. r T r r r T T T r + > Failure rate 1 2 3 4 5 6 FT B 9 10 11 12 (failures per 10* hours) Figure 4 — Estimate and 90% confidence interval for the example Note: The given interval is a confidence interval for the failure rate for the items we have data for. There is no guarantee that items installed in the future will have a failure rate within this interval. MULTI-SAMPLE PROBLEMS In many cases we do not have a homogeneous sample of data, The aggregated data for an item may come from different installations with different operational and environmental conditions, or we may wish to present an “average” failure rate estimate for slightly different items. In these situations we may decide to merge several more or less homogeneous samples, into what we call a multi-sample. The various samples may have different failure rates, and different amounts of data - and thereby different confidence intervals. This is illustrated in Figure 5, p. 27. ©OREDAOREDA-2009 27 Volume 1 — Topside Equipment ‘Sample 2 —— 3 SS k SS Tota! oe ae Soa Se fone Sao We | > Faure rate te BY 5 6 7 8 th 12. (fallres por 10! hours) 4 5 9 10 Figure 5 — Multi-sample problem To merge all the samples, and estimate the “average” failure rate as the total number of failures divided by the aggregated time in service will not always give an adequate result. The ‘confidence’ interval will especially be unrealistically short, as illustrated in Figure 5. We therefore need a more advanced estimation procedure to take care of the multi-sample problem. Below, the so-called OREDA-estimator of the “average” failure rate in a multi-sample situation is presented together with a 90% uncertainty interval. Spjotvoll (1985) gives a rationale for the estimation procedure. The OREDA-estimator is based on the following assumptions: * We have & different samples. A sample may e.g. correspond to an installation (facility), and we may have data from similar items used on é different installations. * In sample no, i we have observed n, failures during a total time in service 7%, for §=1,2,.005k * Sample no. ihas a constant failure rate A, for i=1,2,.... k. * Due to different operational and environmental conditions, the failure rate 2; may vary between the samples. The variation of the failure rate between samples may be modelled by assuming that the failure rate is a random variable with some distribution given by a probability density function 7(A). ‘The mean, or “average” failure rate is then: @= fA-2(A) dA a and the variance is: [a-oy-mayar ) ©OREDAVolume 1 - Topside Equipment 28 OREDA-2009 To calculate the multi-sample OREDA-estimator, the following procedure is used: 1. Calculate an initial estimate 6, of the mean (“average”) failure rate @, by pooling the data: ‘ St Total no. of failures ‘Total time in service 2. Calculate: ' Sot A eo vas, l Ti et Ti 3. Calculate an estimate for o°, a measure of the variation between samples, by: k-1 _ Vk D6, Si-S> M- x §; when greater than 0, elseé? = 4. Calculate the final estimate @ of the mean (“average”) failure rate Oby: In the data tables in Part II of the handbook @ corresponds to the mean (column 4), and SD corresponds to the standard deviation (column 6). @OREDAOREDA-2009 29 Volume 1 — Topside Equipment The /ower and upper “uncertainty” values are given by: viper J m(ayaa=90% Lower Since the distribution a{A) is not known in advance, the following pragmatic approach is used: 6, aA) is assumed to be the probability density function of a Gamma distribution with parameters @and & 7. The parameters @and fare estimated by: e a=2p-O 8. The following formulas are now applied: Lower = 1 Upper = 5 ae where zoosy and zoosy denote the upper 95% and 5% percentiles, respectively, of the 7¢-distribution with v degrees of freedom, see Table 6, page 33. In situations where vis not an integer, an interpolation in the 7*-distribution is performed. Note 1: More detailed analysis of the OREDA data (see Vatn 1993) has indicated that there may be a large variation between installations. The multi-sample OREDA estimator should therefore as a rule be used instead of the n/z estimator which assumes a homogeneous sample. The variation between the samples (installations) is measured by the standard deviation SD. Both the standard deviation and the lower and upper uncertainty values describe the distribution over the failure rate, i.e. 7A). The uncertainty interval should not be mixed up with a confidence interval. A confidence interval decreases with more data, which is not the case for the uncertainty intervals given in this procedure. @OREDAVolume 1 - Topside Equipment 30 OREDA-2009 Note 2: In the case of k= !, the procedure cannot be used. In this case the n/r estimate is given for the mean, and the /ower and upper values should be interpreted as a traditional 90% confidence interval. Note 3: The subdivision of the equipment in detailed taxonomy classes as shown in the data tables will in several cases result in a very low population and/or no. of installations. In these cases it is obvious that the confidence in the data will be very low. The analyst must therefore examine the trade-off between data relevance and population size. We therefore recommend in these cases to use the given data with caution and find additional means to verify the confidence of the data. Some ways this can be done are by using/comparing data from a higher taxonomy level, checking other reliability data sources and/or using expert judgement. Note 4: If no failures are observed for an item, the following approach is used to obtain lower, mean and upper values for “Alll failure modes”: L. Let Ay denote the failure rate estimate (“mean”) one level up in the taxonomy hierarchy. rey . Let r denote the total time in service (operational or calendar) for the item of interest 3. Let a= 1/2 1 =r A 2A, 4. An estimate for the failure rate is now a ee B 5. The standard deviation is given by sp=,|& B 6. A 90% uncertainty interval is given by DOREDAOREDA-2009 31 Volume 1 — Topside Equipment 1 1 _(0.002 1.9 pe pgs ge Alternatively, the term “All failure modes” can be replaced with the failure mode of interest. ESTIMATION OF DEMAND PROBABILITIES If information about “number of demands” is given (see Section “Data Table, Reliability Data”, page 35) it is possible to estimate the demand probability. The demand probability is always related to one specific failure mode, for example a critical fail to start. The maximum likelihood estimator of the demand rate is: a i Prd where 7 is the number of failures with the appropriate failure mode, and ¢ is the number of demands. Note that in the data table presentations the demand probabilities may apparently look different. The reason for this is that in some cases there are registered “demand failures”, but the number of demands is not recorded for one or more inventories. For these inventories, the demand failures are not included in the total number of demand failures for that data table. EsTIMATION OF REPAIR TIMES In OREDA, data on three maintenance related ‘times’ are collected, i.e. down time, active repair time and restoration man-hours. The difference between the first two measures is illustrated in Figure 6. State Rune | Preparation Active patra Raw down | andor delay | maintenance time | “%Y | up | | Up. tine / Down time Up time Time of failure Time Figure 6 — Maintenance times (ISO 14224) Down time includes the calendar time from the equipment is being stopped until it is tested after a repair and ready for being reconnected to its intended service. This parameter is not included in the handbook, only in the OREDA database. ©@OREDA