An advanced methodology for probabilistic risk assessment under limited and uncertain data: Application to offshore accidents

Abstract

The paper proposes an approach for estimating the probabilities of basic causes in an accident scenario using limited and uncertain data from multiple sources. The approach combines expert elicitation and Hierarchical Bayesian Analysis to address the source-to-source and data uncertainties in estimating the frequencies of basic events. The approach is applied to fire/explosion events in Floating Production Storage and Offloading installations. The proposed probabilistic methodology aggregates incidental data collected from specific and similar databases of the leading offshore safety regulators and subjective information provided by expert elicitation on the basic causes of the accident. The basic event frequencies are transformed into posterior probability distributions using Hierarchical Bayesian Analysis to estimate the probability of basic causes of fires/explosions in Floating Production Storage and Offloading installations. The results are compared with the traditional statistical approach of event frequency estimation, and the effectiveness of the proposed methodology is discussed. Applying this methodology is beneficial for better probabilistic modelling of accidents under limited and uncertain data from different sources.