Re-assessing maritime collision risk based on an ontology-Bayesian network model: evidence from China’s coastal waters

Résumé fourni par la source

To address the lack of interoperability among theories and limited knowledge reuse in existing maritime collision literature, this study constructs an ontology-Bayesian network (onto-BN) model that integrates ontology to improve knowledge reuse, and BN to infer and predict maritime collision risk. After analyzing 161 maritime collision cases that occurred in China’s coastal waters, as reported by the China Maritime Safety Administration (CMSA), the main results are as follows: First, a three-tier structure is established, consisting of 1 top class (severity), 4 parent classes (ship factors, management factors, human factors, and external factors), and 19 subclasses (risk factors). Second, a comparison of three BN models indicates that the onto-BN model achieves the highest accuracy in predicting collision risks for the collected cases. Third, it is demonstrated that strict adherence to navigational rules outlined in the Convention on International Regulations for Preventing Collisions at Sea, 1972 (COLREGs 1972) (e.g. maintaining a careful lookout, sailing at safe speeds, and using sound signals) under both normal and extreme conditions can significantly reduce collision risks. These results not only advance the existing knowledge on maritime collisions but also offer valuable insights for enhancing maritime safety and better understanding and compliance with the COLREGs 1972.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.