Study on key causal factors and pathways of fire and explosion accidents in hazardous chemical storage tank area

Abstract

Accidents occurred in hazardous chemical storage areas often have significant impacts and serious consequences. To ensure the safety and health of employees and prevent accidents in hazardous chemical storage areas, it is necessary to explore the key causal factors and critical paths of accidents through certain technical means. Therefore, this paper proposes a research method that combines text mining, association rule mining, and Bayesian networks to conduct in-depth mining and analysis of textual data from cases of hazardous chemical storage tank area fire and explosion accidents (HCSTAFEAs), thereby effectively identifying the causal factors of such accidents and exploring the degree of interaction, importance, and critical paths of the causal factors. First, this paper improved the text mining process by using methods such as grounded theory, text processing, and Chinese word segmentation to mine 60 collected reports, resulting in 68 primary causal factors, 17 secondary causal factors, and 6 tertiary causal factors. Second, the grey relational method was used to analyze the impact of the causal factors, quantitatively determining the importance of each causal factor and further refining them. The Apriori algorithm was subsequently employed to obtain the frequent itemsets and strong association rules of the accident causal factors, and a Bayesian network model was constructed. Through sensitivity analysis and critical path analysis, the key causal factors and critical paths of HCSTAFEAs were identified. The study indicates that five high-sensitivity causal factors—equipment and operation status control defects, equipment maintenance and management defects, unsafe acts, safety management systems and implementation defects, and safety training defects—are the key causal factors of HCSTAFEAs. In addition, three key paths that trigger accidents were obtained: safety management systems and implementation defects → safety training defects → internal supervision defects → operational program defects → unsafe acts; safety management systems and implementation defects → safety training defects → equipment maintenance and management defects → equipment and operation status control defects; and safety management systems and implementation defects → safety training defects → internal supervision defects → operational program defects → equipment and operation status control defects. This paper provides insights into the effective mining and extraction of unstructured accident investigation report textual information and offers a perspective for research on the identification of causal factors and critical paths of accidents in hazardous chemical storage areas based on data-driven thinking.