Handling epistemic uncertainty in the Fault Tree Analysis using interval valued expert information

Abstract

Risk Analysis ( RA) is a meaningful process in every industrial context, particularly referring to major hazard plants. In such sites, input reliability data are generally poor so leading to a partial or incomplete knowledge of the failure process. As a consequence, RA is always affected by the so called epistemic uncertainty which presence makes inappropriate the classical probabilistic approach. Therefore, the present work deals with such a kind of uncertainty in the Fault Tree Analysis (FTA) and proposes a new aggregation rule to combine the interval-valued information supplied by a team of experts about each Basic Event ( BE). The aggregation leads to an interval which bounds are modelled by fuzzy numbers. Successively, such an uncertainty is properly propagated up to the Top Event ( TE) in order to determine the related probability of occurrence. The approach is finally applied and discussed with relation to a Safety Instrumented System (SIS)