Effects of data uncertainty on quantitative risk assessment of an industrial plant under seismic actions

This paper describes a fuzzy set based approach for dealing with uncertainties in the assessment of “NaTech” events triggered by earthquakes. The methodology has been integrated within a general framework developed for the assessment of external hazard factors in industrial sites. The starting point is the Probabilistic Seismic Hazard Assessment (PSHA) at the site considered, in order to evaluate probabilities of occurrence of ground motions of different severity. Fragility curves are then derived for atmospheric storages, corresponding to different damage levels. For both seismic hazard and fragility curves, confidence intervals are derived, and fuzzy numbers are used to model uncertainties. The effects of these uncertainties on Quantitative Risk Assessment (QRA) of local and societal risk indexes, caused by accidental scenarios triggered by seismic events, are evaluated for a case study refinery. The adopted QRA methodology allows for identification and the consequence assessment of all the possible scenarios, including possible domino events. The procedure has been implemented in a GIS-based software tool in order to manage the high number of event sequences that are likely to be generated in large industrial facilities. The developed methodology requires a limited amount of additional data with respect to those used in a conventional QRA.