Accident causation analysis of metal processing plants based on questionnaire and Bayesian network

Abstract

Metal dust and metal waste from the process of metal processing are prone to fire and explosion, resulting in personal injury and property losses. It is necessary to explore the causes of fire and explosion in metal processing plants for the prevention and management of safety accidents. Firstly, this study collected relevant information of fire and explosion accidents in metal processing plants though questionnaire investigation. Then, three experts with authority in this field were chosen, and the experts’ weights were calculated. Secondly, Bayesian networks of metal processing plants were established according to accident information. In addition, questionnaire data and the expert score after fuzzing were applied to serve as probabilistic data for Bayesian networks. Finally, a safety management model was constructed based on the sensitivity analysis in the Bayesian network. It is shown that the probability of human and management errors is the largest in a plant. Compared to other factors, the probability of occurrence of an unsafe state of material is the smallest. Human errors and the ventilation system is not activated are the two factors that have the greatest influence on the accidents. This study provides assistance in improving safety in metal processing plants.