Causation analysis of ship collisions using a TM-FRAM model

• A new TM-FRAM model for multi-dimensional causal analysis of ship collisions. • A novel database of ship collisions invovling new risk influecial factors. • A novel approach to identifying risk factors directly from accident reports using text mining. • Association rule mining of ship collision risk factors using machine learning. Ship collisions pose a significant threat to life and property, presenting a major challenge in maritime safety. Current risk analysis methods have been criticized in terms of a lack of capacity of quantifying the risks of different features and a standardized database reflecting the multidimensional risks of human, mechanical, environmental, and management factors. Additionally, traditional analysis sometimes involves strong assumptions that 1) the established and widely used databases can capture all the essential features of ship collisions and 2) the modelling of ship collision process can be simplified by focusing the analysis on a single causal relationship at once. This paper aims to develop a new approach to enabling multi-dimensional analysis of the causation of ship collisions through the establishment of a new database for ship collisions by innovatively combining Text Mining (TM), Association Rule (AR), and Functional Resonance Analysis Method (FRAM). The new approach enables to construct a risk analysis network based on FRAM, and the model's practicality and effectiveness are validated through expert reviews and case studies. As a result, thirty-eight key risk factors have been successfully identified as per their influence to ship collision incidents, encompassing human error, mechanical failures, adverse environmental conditions, and operational issues. The findings not only offer a new perspective and methodology for ship collision risk analysis, but also enrich the theoretical framework of ship safety management, providing valuable guidance for ensuring ship navigation safety.