Research on Fire Risk Assessment in Machining Workshop Based on Bayesian Network

Abstract

In mechanical processing workshops, fire hazards gravely threaten safety and property. Existing assessment methods fail to comprehensively account for multi-factor interactions. This study innovatively applies the Bayesian network model to this field for the first time. The model, featuring 20 tertiary and 5 secondary nodes, covers hazardous gas storage, operational norms, and equipment conditions. Empirical analysis reveals a 0.16 probability of fire due to improper hazardous gas cylinder storage at the secondary level. Sensitivity analysis shows that "uninspected oxygen and acetylene cylinders upon entry" significantly impacts "non-standard gas welding operations", while hose aging and backfire preventer damage also strongly influence risks. Forward inference identifies key fire-causing factors, with probabilities up to 0.22 at the tertiary level and 0.3 at the secondary level. Reverse inference indicates improper hazardous chemical cylinder storage as the most likely cause during a fire. This study confirms that non-standard operations and equipment aging are pivotal. The Bayesian network's multi-level design and bidirectional inference enable precise risk quantification, offering a scientific basis for safety management and paving the way for dynamic evaluations.