Abstract:

To effectively identify and prevent the risks of chemical fire accidents, a comprehensive research framework was proposed, integrating text mining, SIF model hierarchical analysis, and social network analysis based on rich accident investigation reports in the era of big data. Firstly, the key causes of the accidents were systematically extracted through text preprocessing, term frequency-inverse document frequency (TF-IDF) keyword extraction, and latent dirichlet allocation (LDA) topic modeling, combined with 75 representative chemical fire accident investigation reports from 2000 to 2024. Then, based on the SIF model, the extracted causes were classified into the micro, meso, and macro levels. Subsequently, the accident cause network was constructed using social network analysis methods. Core nodes and key influence paths in the accident cause network were identified through frequency statistics, centrality analysis, and key relationship mining. The research results show that the micro-causes accounted for 56.3%, representing the most crucial factors contributing to chemical fire accidents, with individual factors and environmental equipment risks being dominant. At the meso level, insufficient safety supervision has the highest degree centrality. The key cause path for chemical fire accidents is: inadequate organizational management → insufficient safety supervision → insufficient personal preparation → violation of regulations.

Key words: safety information flow(SIF) model, social network, chemical fire, accident causes, safey management