A novel integrated method for heterogeneity analysis of marine accidents involving different ship types

Abstract

Existing studies have noted differences in risk influential factors (RIFs) across various ship types but often fail to provide a detailed analysis and targeted countermeasures. This study proposes a novel marine accident analysis model that integrates the Complex Network (CN), the Weighted Influence Non-linear Gauge System (WINGS), and the Adversarial Interpretive Structure Model (AISM), to analyse the differences in RIFs for accidents involving different ship types. Firstly, based on 910 marine accident investigation reports covering four major ship types (bulk carriers, container ships, fishing vessels and oil tankers), a RIFs database is established, and the potential relationship between RIFs is mined by association rules. Secondly, Risk Interaction Networks (RINs) for each ship type are constructed, and their topological characteristics are analysed. Subsequently, a dynamic analysis model, named WINGS, is developed to analyse the causal relationship between RIFs from the perspective of dynamic information transmission. Finally, the AISM is established to determine the causal hierarchical relationships among these RIFs. The findings highlight significant differences in the critical RIFs of accidents across different ship types, illustrating the distinct risk profiles and necessitating tailored prevention strategies. This research advances multidimensional factor analysis from static to dynamic, facilitating the development of more tailored preventive measures. The source code is publicly available at: https://github.com/FengYinLeo/CWA-Model . • A novel integrated method for accident heterogeneity analysis across different ship types is proposed. • The proposed method integrates the advantages of CN, WINGS and AISM models. • The proposed model extends the multidimensional factor analysis from static to dynamic. • The key risk influential factors for accidents involving four ship types and their differences are analysed.