New Generic Leak Frequencies for Process Equipment John Spouge DNV Consulting, London SE1 9DE, United Kingdom;
[email protected] (for correspondence) Published online 13 October 2005 in Wiley InterScience (www.interscience.wiley.com). (www.interscience.wiley.com). DOI 10.1002/prs.10100 The likelihood of leaks from process equipment is a vision of safety features, the level of inspection and key input to any quantitative risk assessment (QRA) of maintenance, the permitted inventories of hazardous process plant. This paper describes a new source of materials, and the separation from nearby populations. genericc leak frequencies generi frequencies and revie reviews ws the challe challenges nges in The overall risk management approach is beyond the using it for QRA. scope of this paper, whose focus is on one small but Establ Est ablish ished ed lea leak k fre freque quenci ncies es for ons onshor horee pro proces cess s highly significant input, that is, the frequencies of leaks equipment are poor in quality because they are based from different types of process equipment. on old dat data a sou source rcess of unk unknow nown n pro proven venanc ance, e, jud judg- g- Leak frequencies, which represent the long-term avmentally modified. Recent data from offshore process erag er age e nu numbe mberr of le leak akss pe perr ye year ar of op oper erat atio ion, n, may activi act ivitie tiess pro provid videe a high high-qu -quali ality ty dat data a set set,, whi which ch ap- differentiate between different sizes of leak, from small pears to show that the established frequencies are much to catastrophic. In their simplest form, they refer to too low. Much of this discrepancy arises from the QRA individual equipment items. “Generic” leak frequencies practice of modeling only the most hazardous leak apply to an average of all equipment of this type, as scen sc enar ario ios. s. Re Reco cogn gniz izin ing g th thee pr prac acti tica call di diffi fficu cult lty y of opposed to specific frequencies for particular equipchanging QRA practice to match new data, the present ment manufacturers or operating circumstances. This paper therefore describes a method of analyzing the paper refers to the types of steel equipment used in the new data to obtain leak frequencies for specific mod- process industry to contain liquid and gaseous hydroeled scenarios. carbons. Leaks are divided into three scenarios, allowing an- The role of leak frequencies in a process plant QRA alysts to use frequencies for only those scenarios that is illustrated in Figure 1. Multiplying the generic leak are compatible with their QRA outflow modeling. Stan- frequencies by a plant’s numbers of equipment items dardized leak frequencies have been developed for dif- yields a simple estimate of the likelihood of leaks that ferent types of process equipment, using leak frequency reflects the size and complexity of the plant. This can functions to ensure that consistent, nonzero frequen- be combined with estimates of leak consequences to cies are available for any equipment type and hole size. obtain the overall risk. The dominant sources of leaks The results are consistent with traditional onshore leak can then be identified and appropriate management frequencies, while also being traceable to specific inci- controls adopted. dents den ts amo among ng the mod modern ern hig high-q h-qual uality ity off offsho shore re lea leak k The aim of QRA is to give guidance on managing the data. © 2005 American Institute of Chemical Engineers
Process Saf Prog 24: 249–257, 2005 Keywor Key words: ds: lea leaks, ks, leak fre freque quenci ncies, es, process process equ equip- ip- ment, quantitative risk assessment 1. INTRODUCTION
1.1. The Role of Leak Frequencies Frequencies Modern Mod ern saf safety ety man manage agemen mentt pra practic ctice e comm commonly only makes mak es use of qua quanti ntitat tative ive ris risk k ass assess essmen mentt (QR (QRA) A) to make informed decisions concerning the safety of process plants [1]. These decisions may concern the pro© 2005 American Institute of Chemical Engineers
Process Safety Progress (Vol.24, No.4)
risks of rare events, unlikely to be seen in the direct experience exper ience of most engineers and indivi individual dual plants. Precisely because of their rarity, establishing the frequency of such events is difficult, requiring systematic data collection, covering not only leaks but also the exposed equipment population, over many plants for many man y yea years. rs. Such dat dataa coll collect ection ion is tim time-c e-consu onsumin ming g and thus unusual. Alternative methods such as fault tree tre e ana analys lysis is are pos possibl sible e for pla plantnt-spe specific cific app applica lica-tions, but have not yet delivered generic leak frequencies suitable for routine use in QRA studies. Because they require data for calibration, they are normally used in combination with generic frequencies, rather than as an independent alternative. December 2005 249
Figure 1. Role of leak frequencies in QRA.
1.2. Existing Onshore Leak Frequencies What data collections underpin the generic leak frequencies in current process QRA? Although there are numerous sources available that cite generic leak frequencies [2], few reveal what data underlie them. Most can be traced back to first publication in the 1970s and 1980s. Many have been judgmentally modified to apply to different hole sizes. In almost no case can it be determined what equipment population and hole sizes underlie the leak frequency. In some cases, it is not even clear what industry the values came from, and whether they are based on data or expert judgment. For example, consider what data underlie the frequency of leaks from steel pipes, a key element of most process plants. A typical modern QRA might start with the AIChE/CCPS guide to QRA [1]. The current edition refers to a previous review of sources published in 1989 [2]. Its example of a distillation column used generic frequencies for pipes and vessels from a study of process plants in The Netherlands published in 1982 [3], known by its Dutch acronym COVO. The COVO report gave six separate values for different pipe diameters and leak scenarios, as shown in Table 1, which were based on synthesis of earlier studies, mainly the USNRC’s WASH-1400 [4]. That source is the first in this historical trail to acknowledge some real data: that is, four breaks and four minor leaks from pipes in U.S. nuclear plants during 1972. However, its frequencies were blended with earlier sources, whose origins in the nuclear industry of the 1960s and 1970s are now practically untraceable. Other sources use values of similar quality and age. An authoritative IChemE monograph [5] makes a judgmental synthesis of various sources from 1971–1985. A widely used loss prevention textbook [6] quotes a range of leak frequency data, mainly drawn from WASH-1400. The Netherlands “Purple Book” [7] gives values based on the COVO study, while acknowledging that subsequent reviews indicate a tendency for higher frequencies, but no complete update is yet available. In the case of pipes, it is concluded that the most widely accepted leak frequencies are in reality judgments, resting on a data set of eight leaks in U.S. nuclear plants in 1972, or earlier collections whose size and origin are now unknown. Investigation of other 250 December 2005
types of process equipment yields similar conclusions. This is consistent with an earlier review [8], which concluded that most frequency sources either quoted from other sources or assumed the values. Validation of the frequencies is surprisingly difficult. In part this is because they are so low that most available data are inconclusive. Furthermore, there is no clear definition of what sizes of leaks have been included in the data. Even validation against the original data, where obtained, is problematic. For example, reanalysis of the leaks and equipment population in the WASH-1400 study yields a frequency of 4.3 105 per meter year [8], which greatly exceeds the COVO values that are supposedly based on it. Thus it appears that the judgmental modifications for pipe and leak diameter dominate over the data content in the generic frequency. This may be preferable to the use of unmodified 30-year-old data from a different industry, but it begs the question of why the process industry has not obtained any data of its own during this period. 1.3. Offshore Leak Frequencies Until the 1990s, the same generic leak frequencies were used for process QRAs in both offshore and onshore industries. The inquiry into the Piper Alpha accident in the North Sea [9] recommended that the Health and Safety Executive (HSE) should collect a database of hydrocarbon leaks from offshore installations in the UK Sector, and provide it to operators to support QRA. The resulting hydrocarbon release database (HCRD) has collected all significant releases in the UK Sector since October 1992. In addition, the HSE has estimated the exposed population of equipment items, and from these has determined leak frequencies and size breakdowns for each equipment type. These frequencies were first published in 1997, and most recently updated in 2002 [10], at which time the database contained 2071 leaks. In the future, the offshore industry will have access to the data through the World Wide Web. HCRD is now the primary source of process leak frequencies for offshore QRA [11]. The quality of the HSE offshore data set is exceptionally high, particularly compared to the existing onshore frequencies. For each leak underlying the frequency values, it is possible to establish the hole diameter, the system and equipment type, the hydrocarbon type and pressure, the estimated quantity released, and many other parameters. Figure 2 shows the leak size distribution for all the leaks in the database up to 2001. Convexity in the plot for large hole sizes results from the limited capacity of small-diameter equipment to create large holes. Convexity for small holes in this type of plot would suggest underreporting, which tends to be greater for smaller holes. In this case, the long, relatively straight section of the plot indicates comprehensive reporting. 1.4. Comparison of Offshore and Onshore Leak Frequencies What are the differences between the established onshore and the new offshore frequencies? Figure 3 shows the ratios of the offshore frequencies as analyzed by DNV in 2004 compared to the onshore frequencies Process Safety Progress (Vol.24, No.4)
Table 1. Pipe leak frequencies from COVO study [3].
Pipe Diameter 50 mm
50
150 mm
150 mm
Mode of Failure
Leak Frequency (per Section Hour) (per Meter Year)
Catastrophic rupture Significant leakage Catastrophic rupture Significant leakage Catastrophic rupture Significant leakage
1 109 1 108 3 1010 6 109 1 1010 3 109
8.8 107 8.8 106 2.6 107 5.3 106 8.8 108 2.6 106
Figure 2. Leak size distribution for HSE offshore data.
Figure 3. Ratios between offshore and onshore generic leak frequencies.
that were current in 1993 before the offshore data became available. Some are reduced, but most are increased, sometimes by more than an order of magnitude. The overall effects depend on the equipment types in the study and on the hole sizes used for the comparison. Figure 4 compares results for an example installation, showing the total leak frequency on a base of Process Safety Progress (Vol.24, No.4)
hole size. The overall frequency is similar for both data sets, at about one leak per year, but the slopes of the distributions show two key differences produced by the new offshore values: 1. The most frequent holes are smaller. The overall frequency refers to a minimum hole size of 1 mm December 2005 251
Figure 4. Comparison of overall installation leak frequencies.
Table 2. Causation factors in HSE offshore data [10].
Category Design fault Equipment fault
Operational fault
Procedural fault
Causation Factor — Corrosion/erosion Mechanical defect Material defect Other Incorrectly fitted Improper operation Dropped/impact Left open/opened Other Noncompliance Deficient procedure Other
(based on the data in Figure 2), whereas the traditional onshore values refer to a minimum hole size of 5 mm (a judgment). 2. The largest holes are much more frequent. For hole diameters 100 mm, the new offshore values are more than an order of magnitude higher than the traditional onshore values. In many QRAs, the results are dominated by these large hole sizes. Thus, if the new frequencies are substituted directly into a QRA with no further modification, large changes in the risks can be expected. Although the changes are sensitive to the equipment types and modeling techniques, typically they involve increases of around an order of magnitude in calculated risk. 1.5. Implications of the Different Frequencies The differences in leak frequencies between the onshore and offshore data sets could be interpreted in two entirely different ways:
1. Both sets could be considered valid, with the difference arising from genuinely higher leak frequencies in the offshore industry. 252 December 2005
Instances 321 277 920 76 89 267 495 36 237 81 231 323 34
Category Totals 321 1362
1116
588
2. The offshore data set, being more recent and of higher quality, could also be considered valid for the onshore industry. Are generic leak frequencies higher offshore than onshore? It is a widely held belief in the onshore industry that the more harsh conditions offshore will result in higher leak frequencies. However, the causation factors in the offshore data do not provide support for this interpretation (Table 2). Failure mechanisms associated with the offshore environment (such as salt water corrosion, produced sand erosion, dropped objects, etc.) constitute relatively small proportions of the total and cannot account for the observed order of magnitude difference in frequencies. Given that many installations are under common safety management systems onshore and offshore, it would be expected that where hazards were greater in one environment than another, appropriate management controls would be adopted, with the effect of minimizing any differences between them. This explains why analysts were content to use onshore data for offshore QRAs before offshore data were available. Attempts to make detailed comparisons between Process Safety Progress (Vol.24, No.4)
onshore and offshore leak frequencies for specific equipment types have revealed the poor quality of the onshore values, as discussed above. For example, where the size of the data set is unknown, it is impossible to establish confidence limits on the frequency. For any new onshore QRA, faced with a simple choice between the two data sets, the superior quality of the offshore data would seem to overwhelm concerns about its lack of applicability. However, if the offshore data are adopted for onshore QRA, several further possibilities arise: • The offshore frequencies could be substituted
into onshore QRA with no further modification. This would imply that many existing onshore risks have been significantly underestimated, with severe implications for plants that are to be updated and compared with fixed risk acceptance criteria. • The offshore frequencies could be accepted as the best available estimates of onshore frequencies, but necessitating modifications to the way they are used in onshore QRA. Perhaps existing QRAs incorporated judgmental or fortuitous underestimates of other parameters, such as ignition probabilities, thus counterbalancing the underestimated generic frequencies. This would imply that the whole QRA methodology should be revalidated to use the new data correctly. • Risk estimates based on the offshore frequencies could be accepted as the best available estimates of onshore risks, but changes might be needed in the acceptance criteria. Perhaps strict acceptance criteria have evolved to give sensible decisions when combined with underestimated risks. This would imply that the criteria for decision making should be reviewed to use the new data correctly. Although logical, all these interpretations are some what impractical. In reality, analysts and decision makers wish to use the new offshore data in onshore QRAs, given that these data are the latest and best available, but they do not want to change other modeling parameters or acceptance criteria and they do not want to see major changes to results that have been widely scrutinized and found credible. A more pragmatic solution is therefore required to enable this. 1.6. Challenges in Using Offshore Leak Frequencies The key requirement to enable the offshore leak frequencies to be used for onshore QRA is therefore to obtain frequencies that are compatible with current outflow, ignition, and consequence modeling methodology. Inspection of the HSE data shows that the leak events include many that occurred at zero pressure or whose recorded sizes, pressures, and released quantities indicate that they were quickly isolated. These are not compatible with typical QRA outflow modeling, which assumes continuous flow from the full hole diameter at full system pressure until controlled by emergency shutdown (ESD), blowdown, or inventory Process Safety Progress (Vol.24, No.4)
exhaustion. It appears that this practice of modeling only the most severe leak scenarios accounts for much of the discrepancy between the new offshore data and the older onshore values. In effect, the older values had been adjusted downward to represent only the scenarios normally modeled in onshore QRA. The offshore data, by contrast, are based on a complete collection of all leak events. Therefore, it is necessary to obtain a subset of leak frequencies of scenarios that are consistent with the way onshore QRA is performed in practice. In addition, practical application of the HSE data has revealed the need for several types of adjustments: • Frequencies are available for 89 separate types
and diameters of process equipment (excluding wellhead equipment, drilling equipment, pipelines, and risers). These groups may need to be combined to match the QRA parts counts. • Many of these groups do not have sufficient exposure to show reliable leak frequencies and size distributions. Up to 2003, among the 89 process equipment groups, only 27 have more than 20 leaks and 18 have no leaks at all. These groups may need to be combined to avoid large changes in leak frequency as further data are added. • The Statistics Report [10] gives hole size distributions for seven hole size groups (10 mm, 10 25 mm, 25 50 mm, 50 75 mm, 75 100 mm, 100 mm, and “Not applicable”). This requires adjustments if the QRA needs to model different hole size categories. As a result of the small populations, the frequency is zero for many hole size and equipment type combinations, especially for the larger hole sizes that tend to dominate QRA results. Nonzero frequencies are required to avoid bias in the risk results. The many possibilities for adjustments of this type may result in a wide range of frequencies being derived from the same original data. 1.7. Project Origin The challenges described above are relevant for the offshore industry as well as the onshore industry. Despite the HSE data being accepted as the standard for offshore leak frequencies, different analysts have processed the data in different ways, so that the results of the QRA depend on the consultant carrying out the analysis. To resolve this, the Norwegian operators Statoil and Norsk Hydro established a project to develop standardized leak frequencies. In early 2004, DNV Consulting was commissioned to undertake the work, involving contractors Scandpower and Safetec in the project. The work was completed during 2004 [12] with significant involvement from Statoil and Norsk Hydro. 2. METHOD
2.1. General Approach The project decided to make use of the HSE data from the UK Sector because all three contractors were already using these data, although in different ways. December 2005 253
Available Norwegian data [13] are suitable for validating the approach but, because of lack of equipment populations, it does not give generic frequencies per equipment item. The Norwegian hydrocarbon leak and ignition probability (HCLIP) database is currently being constructed and will eventually provide suitable Nor wegian data. DNV’s method of obtaining leak frequencies from HCRD has three main steps: 1. Grouping data for different types and sizes of equipment, where there is insufficient experience to show significant differences between them. 2. Fitting analytical leak frequency functions to the data, to obtain a smooth variation of leak frequency with equipment and hole size. 3. Splitting the leak frequencies into different leak scenarios, to promote compatibility with different approaches to outflow modeling in the QRA. The leak frequency functions and leak scenarios are described in more detail below. 2.2. Leak Frequency Functions A leak frequency function is an analytical representation of the variation of leak frequency with equipment and hole size. The DNV leak frequency function has been chosen to meet the following general principles: • There should be a smooth variation of leak fre-
quency with hole size and equipment size. • The probability of a given hole size should de-
crease logarithmically up to equipment diameter. • An additional element may be added to represent
ruptures, with a hole size equal to the equipment diameter. This leads to the following general leak frequency function: F d f D d m F rup
for d 1 mm to D , where F (d ) is the frequency (per year) of holes exceeding size d , f (D ) is the function representing the variation of leak frequency with D , D is the equipment diameter (mm), d is the hole diameter (mm), m is the slope parameter, and F rup is the additional rupture frequency (per year). For pipes, flanges, valves, and pig traps, HCRD pro vides data for different equipment size groups. Analysis of these showed significant variations of leak frequency with equipment size for pipes, flanges, and manual valves, and thus the f (D ) term has been defined for these types. The additional rupture frequency F rup and the slope parameter m are assumed to be constants, that is, not to be dependent on equipment size, for any equipment type. 2.3. QRA-Compatible Leak Scenarios To promote compatibility with different approaches to leak outflow modeling in the QRA, the DNV method divides the leaks in HCRD into three main scenarios: 254 December 2005
1. Zero pressure leaks, where the actual pressure inside the equipment is 0.01 barg. This may be because the equipment has a normal operating pressure of zero (such as open drains) or because the equipment has been depressurized for maintenance. 2. Limited leaks, where the equipment is under pressure but the outflow is much less than that from a leak at the operating pressure controlled only by ESD and blowdown. This may be because the leak is isolated locally by human intervention (such as closing an inadvertently opened valve) or by a restriction in the flow from the system inventory (such as leaks of fluid accumulated between pump shaft seals). 3. Full leaks, where the outflow is consistent with or greater than a leak at the operating pressure controlled by ESD and blowdown. This includes: • ESD isolated leaks, presumed to be controlled by
ESD and blowdown of the leaking system. • Late isolated leaks, presumed to be cases where
there is no effective ESD of the leaking system, resulting in a greater outflow. The method of allocating leak records in HCRD into the scenarios is as follows. The initial release rate from the hole is estimated using simplified equations [11], based on the hole size, pressure, and fluid density recorded in HCRD. A range of plausible release quantities is estimated based on the system inventory recorded in HCRD and possible ESD and blowdown responses. Where the recorded release quantity in HCRD is within this range, these are defined as ESD isolated leaks. Late-isolated and limited leaks are cases where the recorded release quantity is respectively above or below this range. As a simple indication of the relative importance of each leak scenario using the methods and criteria above, Figure 5 shows the breakdown of all leaks in HCRD for the period 1992–2003. This shows that nearly 10% of leaks are at zero pressure and 59% are limited leaks. Of the remaining 31% of leaks, 3% are consistent with late isolation, implying that on average ESD has been unavailable on 9% of occasions when it was needed. The breakdown of leaks in HCRD into the scenarios is largely independent of hydrocarbon type, but varies significantly between equipment types and also with hole size. It may therefore be misleading to apply the constant probabilities above for each equipment type and hole size. Instead, DNV allocates each leak in HCRD to a single scenario and then fits the leak frequency functions for each scenario and each equipment type. 3. RESULTS
3.1. Steel Pipes The leak frequency functions obtained by applying the above method to the HCRD records for leaks from steel pipes during October 1992 to March 2003 (inclusive) are [12]: Process Safety Progress (Vol.24, No.4)
Figure 5. Event tree presentation of leak scenarios.
Figure 6. Leak frequencies for 150 mm diameter pipe.
Table 3. Leak frequencies (per meter year) for
• Total leaks , as included in HCRD:
selected hole sizes for 150 mm diameter pipe.
F total 3.7 1051 1000D 1.5d 0.74 3 106 • Full leaks , suitable for modeling as outflow at the
normal operating pressure, controlled by ESD and blowdown: F full 8.0 1061 1000D 1.3d 1.42 • Zero-pressure leaks , occurring with an actual pressure 0.01 barg:
F zero 9.0 106d 0.5 1 106
but not exceeding F total F full • Limited leaks , where the pressure is not zero but
the outflow is much less than that from a leak at the normal operating pressure, controlled by ESD and blowdown: F limited F total F full F zero where F total is the frequency of total leaks (per meter year), F full is the frequency of full leaks (per meter year), F zero is the frequency of zero pressure leaks (per meter year), F limited is the frequency of limited leaks (per meter year), D is the pipe diameter (mm), and d is the hole diameter (mm). Figure 6 illustrates the frequency functions for an Process Safety Progress (Vol.24, No.4)
example 150 mm diameter pipe. Table 3 gives the frequencies for selected leak size ranges. It is notable that the total leak frequency from these results is close to the pipe leak frequency based on the original WASH-1400 data [8], whereas the full leak frequencies are consistent with the more judgmental values quoted in the COVO study. A unique feature of the new frequencies is the ability to access the underlying data set. For example, the frequency of full leaks above is based on 47 events with hole size 1 mm in pipes that are 3- to 11-in. in diameter. The HSE database categorizes the circumstances, causal factors, and consequences of each of these leaks, thus allowing further analysis of unprecedented detail, which as yet has barely started. 3.2. Other Equipment Types Table 4 gives the frequencies of the full leak scenario for different types of process equipment. These are examples of the complete set of generic leak freDecember 2005 255
Table 4. Frequencies of full leaks (per equipment item year) for process equipment.
Equipment Type
Frequency of Full Leaks (1 mm Diameter)
Frequency of Full Leaks (50 mm Diameter)
Steel pipes (2 in.), 1 m length Steel pipes (6 in.), 1 m length Steel pipes (18 in.), 1 m length Flanged joints (2 in.) Flanged joints (6 in.) Flanged joints (18 in.) Manual valves (2 in.) Manual valves (6 in.) Manual valves (18 in.) Actuated valves (6 in.) (nonpipeline) Instrument (0.5 in.) Process vessel Centrifugal pump Reciprocating pump Centrifugal compressor Reciprocating compressor Heat exchanger (h/c in shell) Heat exchanger (h/c in tube) Heat exchanger (plate) Heat exchanger (air cooled) Filter
5.7E-05 2.0E-05 1.1E-05 3.2E-05 4.3E-05 1.2E-04 1.4E-05 4.8E-05 2.2E-04 2.6E-04 2.3E-04 5.0E-04 1.8E-03 3.7E-03 2.0E-03 2.7E-02 1.4E-03 1.0E-03 6.0E-03 1.2E-03 8.9E-04
0.0E00 7.7E-08 4.2E-08 0.0E00 3.6E-07 1.1E-06 0.0E00 4.9E-07 2.3E-06 1.9E-06 0.0E00 1.1E-04 2.4E-05 5.2E-04 2.0E-06 1.1E-05 1.3E-04 4.9E-05 3.6E-04 6.9E-05 6.4E-06
quency results that now form the standardized leak frequencies for offshore projects, and are also considered suitable for onshore QRA studies. The overall results of using these new frequencies for an example installation are included in Figure 4. This shows that the full leaks are roughly an order of magnitude less likely than the unmodified total leaks and are consistent with the traditional onshore values for large hole sizes. 3.3. Sensitivity Tests Sensitivity tests have been conducted to identify the main sources of uncertainty in the generic frequencies, focusing in particular on the full leak scenario. Although there is no evidence of systematic underestimation of the release quantities in HCRD, it is an unavoidable limitation of the approach that the results would be very sensitive to any such bias. The results are also sensitive to the treatment of cases where the system inventory was not recorded, and so the importance of this parameter should be noted in any future data collection. The results are very sensitive to the assumed ranges of isolation and blowdown times, and more realistic modeling of these aspects would be desirable in future work. CONCLUSIONS
The hydrocarbon release database collected by the HSE in the UK offshore industry contains data of a quality and quantity that far surpasses any previous leak data in the process industry. Although becoming the standard for offshore industry, it has rarely been used in onshore QRAs because it would tend to give much higher risks than the established but largely judgmental onshore leak frequencies. The approach de256 December 2005
scribed here solves this problem by dividing leaks into three scenarios, allowing analysts to use frequencies for only those scenarios that are compatible with their QRA outflow modeling. Standardized leak frequencies have been developed for different types of process equipment, using leak frequency functions to ensure that consistent, nonzero frequencies are available for any equipment type and hole size. The results are consistent with traditional onshore leak frequencies, while also being traceable to specific incidents among the modern high-quality offshore leak data. Despite the arguments for similarity in this paper, it remains undesirable to use offshore data for onshore QRA. This is adopted only because of the poor quality of available onshore frequency sources. A key longterm aim for the onshore industry should therefore be to gather leak frequency data of a quality comparable to the HSE offshore data. ACKNOWLEDGMENTS
The work reported in this paper is based on data collected by the Health and Safety Executive, and was funded by Det Norske Veritas (DNV), Statoil and Norsk Hydro. The author acknowledges their kind support; in particular Stine Musæus, Brian Bain, Jens Michael Brandstorp, and Jan Pappas. The author also thanks Art Dowell of Rohm & Haas and John Covan of Sandia National Laboratories, who reviewed the paper and provided helpful comments. Views expressed are those of the author and not necessarily those of DNV. LITERATURE CITED
1. Center for Chemical Process Safety (CCPS), Guidelines for chemical process quantitative risk analysis Process Safety Progress (Vol.24, No.4)
2. 3.
4. 5. 6. 7.
(2nd ed.), American Institute of Chemical Engineers, New York, 2000. CCPS, Guidelines for process equipment reliability data, American Institute of Chemical Engineers, New York, 1989. Rijnmond Public Authority, A risk analysis of six potentially hazardous industrial objects in the Rijnmond Area—A pilot study, COVO Commission, Reidel Publishing, Dordrecht, The Netherlands, 1982. U.S. Nuclear Regulatory Commission (USNRC), Reactor safety study, NUREG-75/014, WASH-1400, USNRC, Washington, DC, 1975. A.W. Cox, F.P. Lees, and M.L. Ang, Classification of hazardous locations, Institution of Chemical Engineers, Rugby, UK, 1990. F.P. Lees, Loss prevention in the process industries (2nd ed.), Butterworth-Heinemann, Oxford, UK, 1996. Committee for thePrevention of Disasters(CPR),Guidelines for quantitative risk assessment (Purple Book), CPR 18E, CPR, The Hague, The Netherlands, 1999.
Process Safety Progress (Vol.24, No.4)
8. S.H. Bush, Reliability of piping in light water reactors, Proc Symp on Application of Reliability Technology to Nuclear Power Plants, IAEA-SM-218/11, International Atomic Energy Agency, 1978. 9. The Hon. Lord Cullen, The public inquiry into the Piper Alpha disaster, Department of Energy, London, UK, 1990. 10. Health and Safety Executive (HSE), Offshore hydrocarbon release statistics 2001, HID Statistics Report, HSR 2001 002, HSE, Bootle, UK, 2002. 11. J.R. Spouge, A guide to quantitative risk assessment for offshore installations, CMPT 99/100a, Centre for Marine and Petroleum Technology, UK, www. mtd.org.uk 12. Det Norske Veritas (DNV), Offshore QRA standardised hydrocarbon Leak frequencies, DNV Report 2004-0869 to Statoil ASA and Norsk Hydro, 2004. 13. Petroleum Safety Authority (PSA), Trends in risk levels—Norwegian continental shelf, Phase 3, PSA, Norway, 2003.
December 2005 257
