Modelling ship collision risk based on the statistical analysis of historical data: A case study in Hong Kong waters

Abstract Collision, as a common type of ship accidents, leads to serious property loss and personal injury. In this paper, a new framework of quantitative risk assessment is proposed by quantifying the probability and the corresponding consequence based on the historical accident data. Firstly, the consequences of ship collisions are quantified and classified using an equivalent consequence method. Secondly, a decision tree model is established to analyse the impact of ship attributes on the collision consequences. The main ship attributes contributing to collision are determined, based on which, a BP neural network model is developed to estimate the probabilities of the different consequences. Thirdly, the collision risk is predicted by integrating the collision probabilities with the corresponding consequences. Fourthly, a case study in Hong Kong waters is investigated and the results are compared with the available references to validate the proposed framework. The new model can be used to assess present risks to plan preventive measures for the potential collision accidents.