Bayesian-Based Collision Risk Assessment for Ships in Navigation Tunnels

Navigation tunnels are special waterway structures that help ships pass through mountains, connect rivers, and improve sailing conditions. At present, the Goupitan Navigation Tunnel is the only one in operation in China, while most other projects are still in the design or testing stage. To study ship safety in such tunnels, this paper develops a framework for collision risk assessment using Bayesian networks. The framework combines uncertainty modeling with full-scale ship experiments and includes three main steps: model building, quantification, and validation. In the building stage, risk factors are identified from five aspects: tunnel geometry, navigation state, human factors, environmental conditions, and infrastructure conditions. A Bayesian network model is then created and improved with expert knowledge and experimental data from the Goupitan Navigation Tunnel. In the quantification stage, conditional probabilities are used to describe the relationships between nodes, and prior probabilities are updated with experimental data. In the validation phase, the model was examined through axiom testing, sensitivity analysis, and extreme scenario analysis. The results demonstrate that the constructed Bayesian Network can effectively evaluate collision risks under different scenarios and identify key nodes. This study provides theoretical support and practical insights for ship safety management and collision risk prevention in navigation tunnels.