Intelligent Use of Sparse Data for Risk Assessment of Rare Events, a Case Study

Abstract

Risk assessments of offshore operations in the North Sea are primarily concerned with risk from major accidents. North Sea offshore operations have increased rapidly from its early start approximately twenty years ago, and there are relatively few cases of major accidents to form a basis for risk assessments. This is often the case for situations in which Probabilistic Risk Assessments are used.

Consequently an examination of intelligent use of sparse data for such situations should be useful beyond the actual area discussed in the paper, the offshore operations. Major accidents to a semi-submersible drilling platform from marine hazards present a typical case where risk assessment is needed in a critical area, based on sparse data.

The principle presented in this paper is an alternative to what is often used in PRAs, where all existing experience data are collected, pooled and used as the basis for an overall frequency assessment. The approach uses a combination of historical frequencies and subjective evaluations. Bayesian uncertainty analysis is indicated as the means to compare the suggested approach with a traditional approach.