Risk assessment of ship navigation using Bayesian learning

Abstract

Risk has a random uncertainty. Risks associated with a ship navigation at sea are analyzed to solve the problem of uncertainty and a developing method is applied to be feasible to work out. Based on Bayes' point estimate and Bayesian learning to estimate the traffic accidents related to ship navigation, an analysis model is established for the quantitative risk assessment (QRA) of the vessel traffic system at sea. After the analysis on occurrence likelihood of the accidents related to ship traffic, a structure on the basis of Bayesian networks is developed to obtain the QRA of their relative risks. QRA is also put forward after analyzing the features and situations of the vessel traffic system and identifying the corresponding feature including characteristics of those hazards. The risk distributions of ship navigation are described and results are presented on QRA in relation to various features by using this method. This method, verified in the cases of QRA, turns out to be feasible by the use of machine learning.