Bayesian network-based failure analysis of fire safety barriers in floating LNG facility

Fire accidents on floating LNG (FLNG) processing facilities are initiated by the uncontrolled gas leaks and then evolved through the successive failure of relevant safety barriers (SBs).This paper develops a methodology to perform comprehensive failure analysis of fire SBs and corresponding consequences in FLNG system using a Bayesian network (BN).Two BN models were constructed and compared, the independent BN and the dependent BN. The BN models were mapped from Event Tree-Fault Tree (ET-FT) diagram.The inter-dependency between SBs and uncertain causal relationships was completely considered in the dependent BN model. The dependent model also investigated the dependency of all SBs on the release prevention barrier (RPB) as the critical and initial barrier of the fire safety system. The probabilities of SBs failure obtained using the dependent BN were less than that of independent BN values. This could be ascribed to the influence of SBs inter-dependency and the model uncertainty due to uncertain relation relationships between managerial and organization factors (MOFs) and functional failures. Among SBs, it was found that the RPB has the highest failure probability. The diagnostic analysis further identified the MOFs as the most influencing factors contributing to SBs failure. The predictive analysis revealed that the large fire with major damage has a higher probability to occur in comparing with other consequences. Additionally, the BN was extended to estimate the risk levels of consequences by integrating the accident probability, accident severity, and escape, evacuation, and rescue (EER) process reliability. The findings indicated that the EER operations have a major influence on the risk level and suggested to be considered for reliable risk level assessment rather than traditional risk matrix.