Research Of Risk Assessment For Ship-bridge Collision Based On Bayesian Network

As China’s Comprehensive National Power(CNP)increases, the transportation industry also enjoys a rapid development. In order to boost economy in some regional areas, more and more bridges are raised up over rivers and sea. Despite economy development these bridges have promoted, they have meanwhile caused impacts on the development of water transportation industry. In the recent years, the number of bridges gradually has been growing, the size of bridges has been increasing, and the water traffic has become noticeably denser. These changes have induced in remarkable risks to both ships and bridges. Thus, it is more than important to study the risks where bridges are established over water traffics. To prevent unexpected accidents from happening, prealarming system must be set and activated when ships are coming to bridges, and after case measures must be set and activated when ships are coming to bridges, and after case measures must be conducted. Improving the safety when ships pass by bridges is by all means of great significance.Based on existing studies, this thesis used orbit intersecting coupling system. That is, through analyzing historical incidents of ships hitting on bridges, to perform a comprehensive analysis against the four factors: human, ships, environment and management, and conclude the main determiner of the collision.With the use of causal relationships, this thesis established a Bayesian network structure to serve the purpose of ship-bridge collision risk analysis. The methodologies of MDL measurement and local search are also carried out to best optimize the Bayesian network structure. SPSS will be applied to carry out K-S test against the conditional possibilities from expert consultant, as well as Mann-Whitney U test against two independent samples. Only if the given possibilities are proved effective, they will be averaged to form a node distribution of the Bayesian network which is the final Bayesian network model.Next, the HUGIN software will be utilised to test the ship-bridge collision model, and the Bayesian network will be used to infer the pretest possibility and posttest possibility. Then, the most dangerous factor will be identified and all the factors will be sorted out in order according to their risk coefficients. Moreover, the most serious incident causing chain will be identified among all and forecast risks. Lastly, preventive measures and recommendations around human factors. The Bayesian network shipbridge collision model not only can effectively reduce the incidents of ship-bridge collision, but also enables scientific and humanistic controls and preventable measures.