Frequency data and modification factors used in QRA studies

The source data for QRA’s is important to assure meaningful risk assessment results, particularly when the result is to be compared against quantitative risk acceptance criteria. The author’s company is one of the largest global QRA providers and we have concluded that the UK HSE Hydrocarbon Release Database (HCRD) provides the basis for the best leak frequency data as it offers complete leak data collection in a systematic manner, against a known population of equipment and facilities in the UK sector of the North Sea for which there is an accurate parts count estimate. The LEAK program is described. It screens HCRD to remove leak events not associated with full operations inventories and flows (e.g. when isolated for maintenance) and further uses a distribution function that permits frequencies for any arbitrary hole size range to be determined (e.g. 25 mm leaks, 50 mm leaks, full-bore ruptures). An important factor is that leak frequency data is not stationary, offshore operators have improved their control of leak events and the HCRD shows a declining leak rate over time.DNV often uses frequency modification techniques, termed MOR - Modification of Risk. This paper reviews 4 methods developed by the company internationally. These are the Manager Method, the API 581 method, a barrier based method, and a proprietary management system based method. These all permit localization of UK North Sea data to apply to other facilities (onshore or offshore) and with different management systems and mechanical integrity programs.Overall, localized data using MOR is considered more accurate than direct use of UK North Sea data, however validation is an issue. There are no direct comparisons that compare leak statistics over a sufficiently long period with static management systems and integrity programs. Thus MOR techniques remain judgment based approaches, but transparent in methodology and assumptions. The barrier based modification technique is the most directly verifiable of the four MOR methods presented.