A hybrid Monte Carlo and possibilistic approach to estimate non-suppression probability in fire probabilistic safety analysis

Abstract

In Fire Probabilistic Safety Analysis (FPSA), the non-suppression probability (that quantifies the likelihood that the installed protection system fails to protect the target from fire) is typically estimated using predefined detection-suppression event trees, that are expected to cover uncertainties with conservatism. In this study, a hybrid Monte Carlo (MC) and possibilistic approach is proposed for uncertainty propagation and effective quantification of a protection system non-suppression probability. In particular, aleatory uncertainty is represented by probabilistic distributions and treated by MC sampling, whereas, epistemic uncertainty of human behavior by means of possibility distributions. The approach is applied to a switchgear room of a Nuclear Power Plant (NPP). Uncertain responsiveness of the fire protection system is integrated into a detection-suppression event tree, allowing for a clearer modeling interpretation and a more accurate failure probability estimate.