Risk assessment of marine accidents with Fuzzy Bayesian Networks and causal analysis

Marine shipping is a high-risk mode of transportation in a complex environment. Most risk analyses of marine shipping are based on historical accident statistics or expert opinions rather than available evidences that reflect more realistic scenarios. In this paper, an evidence-based Fuzzy Bayesian Network approach was proposed to construct probabilistic models of marine accidents. Marine accident reports from 2001 to 2020 were utilized to build causal networks from a systematic perspective. The quantitative relationships of contributing factors in the accident networks were evaluated by experts with different backgrounds. The reliability of the Fuzzy Bayesian Networks of accidents was also verified by three axioms. The results of the sensitivity analysis revealed that distraction, heavy weather, inadequate safety/risk awareness, and maintenance failure are closely related to a majority of high-consequence marine accidents. In particular, the occurrence of heavy weather has a relatively higher impact on contact, flooding/foundering, and grounding. Overall, this paper contributes to a practical direction of ocean and coastal shipping safety management strategies and the quantitative effects of safety measures considering the allocation of limited resources.