Skip to main content

Safety in numbers

Published by , Senior Editor
Hydrocarbon Engineering,


As an inherently hazardous industry, the global oil and gas sector invests heavily in new technology, R&D and improved processes and procedures to reduce the likelihood and consequence of accidents. It is understandably a complex task to assess the relative value of such investments to achieve an appropriate balance of financial expenditure with risk reduction.

While all hazards need to be considered in a full risk assessment, the focus of attention is often the hydrocarbon loss of containment events. Hence, one of the key inputs to a risk assessment is the estimate of frequency of accidental releases of different sizes. This estimate is based on the analysis of historical incidents.

The UK’s Hydrocarbon Release Database (HCRD)1,2,3 has captured the details of almost 5000 leaks on UK North Sea offshore installations from 1992 to the present day. It is the most used source of leak frequency data in the UK and is also applied to many other parts of the world, as well as being used as guidance for onshore process plants and refineries.

On behalf of the International Association of Oil and Gas Producers (IOGP), DNV GL has developed and revised correlations for leak frequency and hole size distribution. This analysis indicates a significant reduction in leak frequency, which has direct implications for risk-based design and operational decision-making. Alternative correlations produced by organisations in the Norwegian offshore industry show a similar decrease.

Background

Risk from hydrocarbon loss of containment events is directly proportional to leak frequency. Hence, the leak frequency estimate will directly affect risk-based decisions on the provision of safety barriers and the costs of constructing and operating facilities.

With such a high reliance on the use of the HCRD as a source, it is important that the reporting, recording, collation and interpretation of data is thorough and technically sound. There are currently initiatives in place, or planned to further improve the data gathering and thereby the accuracy of the estimates produced.

Recent work by DNV GL identified some areas for improvement and a revised model for calculating suitable estimates for process equipment leak frequencies has been developed.4

In recent years, hydrocarbon leak events in the Norwegian Continental Shelf (NCS) have been analysed to provide an alternative set of correlations known as the Process Leak for Offshore Installations Frequency Assessment Model (PLOFAM).5 Both this and the most recent analysis of the HCRD indicate significantly lower frequencies than those used previously. Use of these new correlations will reduce the calculated risks when applied in the analysis of a given installation.

In September 2019, the IOGP revised its guidance on the calculation of process leak frequencies and used the analysis of the HCRD carried out by DNV GL as the primary source.6 Details and parameters for the PLOFAM model are also included in the IOGP report for use in risk assessment of Norwegian offshore installations.

The UK database

Since the Piper Alpha disaster in 1988 and the subsequent enquiry led by Lord Cullen,7 the UK offshore industry has been required to provide data on loss of containment incidents. This was achieved by systematically recording incidents that occurred on installations in UK waters. This was intended to provide a structured set of data that would identify trends and could be used as a source of information for risk analyses. The UK HSE set up a system to gather these data alongside the population of equipment in which these incidents were occurring, so that frequencies could be derived.

The latest publicly available information from the HCRD – until 31 December 2015 – gives details of 4656 incidents. Provisional data for 2016 and 2017 has also been made available to a joint industry project (JIP) for research purposes.

The data collected up until 2015 has 118 fields of information for each record. This includes details on the installation, the area where the release took place, the cause of the leak, how it was detected, information on ignition, and any emergency action which followed as a result. While all of these are important and merit detailed analysis, relatively few fields are required to establish frequency distributions for various types of equipment.

Figure 1 shows the variation in the number of incidents reported on a year-by-year basis. This shows a downward trend indicating that it may now be appropriate to use revised process leak frequency correlations used in risk assessments. However, the population of process equipment should also be taken into consideration to establish an accurate model.


Figure 1. Number of reported incidents per year in the HCRD.


Between 1992 and 1994, the HSE collected systems and equipment population data for UK Continental Shelf (UKCS) installations. This provided data on the total population of 120 categories of process equipment in the UKCS. In 2002, an online system was introduced to enable operators to submit and update their equipment population data.

Some anomalies were observed in the population dataset and in 2016 DNV GL, along with the HSE and OGUK, carried out an exercise to derive better estimates. This included identifying missing data and adding estimates based on comparisons with equivalent process systems and installations. The dataset produced and available on the HSE website is one of the main inputs to the derivation of the IOGP correlations. As a result of legislation introduced by the European Union in 2014,8 the requirements for reporting have changed. The new report of an oil and gas incident (ROGI) form has a mixture of compulsory and voluntary information.9 However, some data considered necessary by risk analysts, such as the equipment type and the equivalent hole diameter, fell into the voluntary category meaning that it was less likely to be supplied than before. This has led to a reduction in the completeness and quality of the data.

Norwegian database

Previously, risk assessment of offshore installations in Norwegian waters used the HCRD as the leak frequency data source. However, a recent study involving several operators and consulting organisations in Norway, adopted an alternative approach for deriving frequency estimates.5 This has resulted in the creation of a model named PLOFAM.

Although, the model is mainly validated against available NCS leak data, it also takes account of information from the HCRD. Leak frequencies are based on the period from 2006 – 2017 while the whole period (2001 – 2017) is used for the relative leak rate distribution.

Information has been gathered on 260 incidents recorded at installations located on the NCS in the period 2001 – 2017 inclusive.

Data was obtained from equipment counts compiled for use in quantitative risk analyses for 79 installations, which were in operation in the NCS at some point during the period 2001 – 2017. For a further 25 installations, where equipment counts were not available, estimates were made by comparing them with similar installations.

Derivation of leak frequencies

Not all incidents reported are considered relevant to a quantitative risk assessment (QRA) study. In addition, the population data needs to match the incident data. Therefore, incidents relating to installations not included in the population dataset are excluded from the calculation, as are systems such as water treatment, flaring/venting systems or utilities. Incidents with hole sizes less than 1 mm dia. were also excluded, along with incidents occurring in depressurised process systems.

Taking all criteria into account, the number of incidents from the HCRD dataset was reduced from the original finding of 4563 to 1964. Similar, but not identical, criteria were applied to the NCS dataset resulting in 217 relevant incidents for use in the PLOFAM model.

The frequency of leaks greater than or equal to 1 mm can generally be determined by dividing the estimated number of leaks in that period by the population, measured in equipment-years, to which it applies. Both studies based the frequencies on experience from 2006 onwards, although in the DNV GL analysis this period was extended backwards for equipment types which had limited numbers of incidents.

The available data for a given type of equipment can be arranged in order of hole size and a plot of the frequency of exceeding that hole size produced. An example of this is shown as the blue line in Figure 2. A curve fitting process is then applied to produce a mathematical function which gives a good approximation to the data. This is represented by the black curve in Figure 2.


Figure 2. Historical data and corresponding mathematical correlation.


For some equipment types, such as flanges, there is sufficient data to be able to distinguish between equipment of different sizes, i.e. the correlation for a 3 in. flange will be different from that for a 12 in. flange.

IOGP guidance provides details of the parameters along with graphical and tabulated representations for 24 equipment types. As an example, Figure 3 shows the graph for flange joints, which indicates the effect of equipment size.


Figure 3. Frequency exceedance curves for flanges using the IOGP correlations.


The frequency of leaks within a certain hole size range, e.g. between 10 mm and 25 mm, is evaluated by subtracting the value exceeding the upper range from the value exceeding the lower range.

The correlations can also be used to generate tables such as the example shown in Table 1.


Table 1. Leak frequency data for different sizes of flange.

Effect on using the new leak frequency models

Both the IOGP and PLOFAM models give estimates of significantly lower frequencies than previously used. In the case of the IOGP formulation, the values are approximately 60% lower.

Given that the risk is directly proportional to the leak frequency, the calculated risk will fall by a similar amount. As such, remedial measures, which were previously judged to be required, may now be considered as no longer justified. The main factors for the reduction in the IOGP estimates are:

  • The reduction in the annual number of incidents reported.
  • The decision to base frequencies on the last 10 years of available data.
  • A reassessment of the incidents to be included in the analysis.
  • A reassessment of the equipment population data.
  • The PLOFAM model produces estimates for leak frequencies, which are lower than the IOGP correlations and this is particularly the case for larger hole sizes. This implies a bigger reduction relative to previous estimates compared with those obtained from the IOGP model.

Towards improved accuracy and value

The collection of data in the UK and Norway should continue in order to gain more statistical significance for the various equipment types. It is also necessary to confirm or otherwise that the downward trend continues. Some observers believe that the current level of reporting marks a low point in the frequencies of events and that a combination of ageing and the effects of the downturn in the oil industry on maintenance levels will lead to an increase in the coming years.

A review of the population data in the UK is required to ensure a robust database is available to accurately assess frequency of leaks. The UK industry has continued to work to improve the estimates of the equipment population. In 2020, a system will be implemented which requires operators to review the equipment counts for their installations when updating their safety case documents: a five-yearly process. This should ensure an improved estimate of the equipment count and enable refinements to be made to the parameters used in the leak frequency correlations.

The quality of the incident data collected needs to be actively managed with improvements in the UK reporting format and encouragement to complete all data fields as accurately as is practical.

These two initiatives will improve and maintain the quality of the data gathered in the UK in future years, and thereby make the complete dataset more valuable.

Findings from OGUK’s Health and Safety Report 201910 revealed that hydrocarbon releases in the ‘major’ category, whilst reduced since 2012, have since increased to four in 2018. Continued industry efforts to drive concerted action in this area are being steered by OGUK in partnership with Step Change in Safety.


Written by Brian Bain, Zoe Wattis and Carol Humphreys, DNV GL – Oil & Gas.

Note

This article is an abridged copy of a paper that was due to be presented at AIChE Spring Meeting, Houston, US, 30 March – 1 April 2020 (postponed to August 2020).

References

  1. HSE, Spreadsheet entitled ‘Offshore Hydrocarbon Releases 1992 – 2016,’ http://www.hse.gov.uk/offshore/statistics.htm
  2. HSE Spreadsheet entitled ‘Offshore Hydrocarbon Releases 2015 – 2016,’ http://www.hse.gov.uk/offshore/statistics.htm
  3. HSE x Spreadsheet entitled ‘Offshore Hydrocarbon Population Data 2015 – 2016,’ http://www.hse.gov.uk/offshore/statistics.htm
  4. BAIN, B, ‘Updated Leak Frequency Modelling Based on the UK Hydrocarbon Release Database’, Hazards 27 Symposium, (2017).
  5. Lloyd’s Register Consulting, Process Leak for Offshore Installations Frequency Assessment model – PLOFAM (2). Report No. 1007566/R1 Rev: Final, (December 2018).
  6. IOGP, Risk Assessment Data Directory, ‘Process Release Frequencies’, Report 434-01, (September 2019).
  7. CULLEN, The Hon. Lord W. Douglas. The public inquiry into the Piper Alpha disaster. London: H.M. Stationery, (1990).
  8. European Union, ‘Commission Implementing Regulation (EU)’ No. 1112/2014, (2014).
  9. Offshore Safety Directive Regulator, Report of an Oil and Gas Incident (ROGI) Form, https://www.hse.gov.uk/osdr/assets/docs/rogi.pdf

Read the article online at: https://www.hydrocarbonengineering.com/special-reports/08092020/safety-in-numbers/

You might also like

TotalEnergies and SINOPEC join forces to produce SAF

TotalEnergies and China Petroleum and Chemical Corp. (SINOPEC) have signed a Heads of Agreement (HoA) to jointly develop a sustainable aviation fuel (SAF) production unit at a SINOPEC's refinery in China.

 
 

Embed article link: (copy the HTML code below):


 

This article has been tagged under the following:

Downstream news