Abstract This paper proposes a Bayesian Network-based framework for risk assessment and probability estimation of vessel-platform allision accidents, using a novel technique that derives probabilities from incidental data. A dataset of 557 allision incidents collected from multiple open source agencies is analysed to identify causation patterns. Basic causes could only be determined for 375 incidents, with supply vessels involved in 61% of cases. Statistical analysis revealed that vessel type and the month of occurrences are significantly associated, and most incidents arose during cargo transfer operations. Fixed installation accounted for the majority of allisions with moving vessels, and human error emerged as the leading contributor (30%). Building on these insights, a Bayesian Network model is developed incorporating 42 identified causes, three causal factors and four consequence levels. Using a recent probabilistic approach, probabilities of basic causes are derived from annual allision occurrence rates. The BN model is then applied to predict annual allision probabilities and to conduct sensitivity analyses. Results show that weather-related causes and misalignment errors exert the strongest influences on accident probabilities. The methodology is transparent and holistic in providing better discernment of the causation probability of allision accidents.
Bhardwaj et al. (Sat,) studied this question.