A method for assisting the accident consequence prediction and cause investigation in petrochemical industries based on natural language processing technology

This study mines petrochemical risk-analysis text (HAZOP historical records) to assist accident consequence prediction and cause investigation. It uses topic modeling (LDA) to cluster causes and consequences, part‑of‑speech tagging to identify representative phrases, and association rule mining (Apriori) to derive likely cause–consequence linkages, presented through a visual interface. For PSM incident investigation, the method supports faster and more consistent reasoning about how abnormal situations can evolve and which consequences are most plausible given observed causes.