Navigation Risk estimation using a modified Bayesian Network modeling-a case study in Taiwan

Abstract

Frequency of the shipping accidents in Taiwan has been increasing since 2013. Navigation Risk estimation using a modified Bayesian Network (BN) making use of the shipping accident data between 2014 and 2019 is conducted. Parameters based on Ministry of Transportation and Communications (MOTC) marine accident database are treated as the parent and child nodes, namely, Ship Age, Flag, Ship Type, Gross Tonnage, Accident Type, Accident Severity, Sea State and Location. The information of such knobs is compiled and forms the basis for prior and conditional probability calculations. Considering the consistency of the condition probability entries for child nodes lacking data, a mapping process contemplating the states from parent nodes is proposed. Rationality of the BN model is validated by the sensitivity analysis of the Navigation Risk outcomes and Accident Frequency and Accident Severity comparisons with the MOTC figures. General Cargoes, Bulk Carriers and Containers under 10,000 GT tend to have higher risk of accidents whereas Containers and General Cargoes over 15,000 GT are prone to encounter mishaps. Grounding, Collision and Fire/Explosion should deserve attention worldwide. The proposed BN demonstrates the feature of proactive safety analysis based on scenario analysis carrying out Navigation Risk predictions considering vessel characteristics and environmental conditions.