On the causation analysis for hazards involved in the engine room fire-fighting system by integrating STPA and BN

Abstract

A novel methodology is proposed in this study to investigate the hazards involved in the engine room fire-fighting system. In the proposed methodology, the fuzzy theory is applied to bridge the superiority of the systems theoretic process analysis (STPA) and Bayesian network (BN). The engine room fire-fighting system is initially audited within the framework of STPA qualitatively to identify unsafe control actions, and then, the causations contributing to these actions are determined, which are subsequently considered as the root nodes of the BN. To implementing Bayesian inference, the expert elicitation within fuzzy theory is employed to determine the probability distribution of the root nodes, and the conditional probability tables for intermediate nodes are obtained by leaky Noisy-OR function, based on which the developed BN is simulated. Finally, the results from BN simulation are interpretation from the perspectives of brief propagation analysis, sensitivity analysis and influence strength analysis. The proposed methodology can be applied to audit the integrity of engine room fire-fighting system to eliminate the potential causations contributing to engine room fire.