Insight process safety of a hydrogen turbine supply system: A comprehensive dynamic risk assessment using a fuzzy Bayesian network

Abstract

This study conducts a comprehensive risk assessment for a selected hydrogen supply system within a pilot hydrogen power plant. The assessment begins with hazard identification through Hazard and Operability (HAZOP) analysis, followed by quantification using bow-tie analysis and consequence modeling to evaluate the impacts of credible scenarios resulting from random failures. Furthermore, a fuzzy Bayesian network is employed to address uncertainties inherent in the quantitative approaches and to account for event combinations and dependencies. Sensitivity analysis is subsequently conducted to identify critical events. Utilizing three quantitative approaches, fault tree analysis, the Bayesian network, without and with evidence updating, the probability of the top event is determined to be 4.52 × 10−2/year, 4.50 × 10−2/year, and 4.70 × 10−2/year, respectively. Additionally, the estimated firezone resulting from the explosion's 300 mbar overpressure was 12 m. The findings highlight the significant role of pressure regulator valve and pressure transmitter failures, with posterior probabilities of 0.17/year and 1.90 × 10−3/year, in ensuring the system's safety.