A Quantitative Evaluation Method for the Safe Operating Status of Gas (Non-Condensing Natural Gas) Boilers Using Bert Pretrained Model

With increasingly stringent safety and environmental requirements, the operational safety of industrial gas-fired boilers has become a critical concern. This study proposes an intelligent and quantitative evaluation method for assessing the operational safety risks of natural gas-fired boilers. Firstly, the method establishes a comprehensive safety operation index system for gas-fired boilers, covering manufacturing, installation, operation, inspection, management, and supervision, based on national and industry standards. Then, it determines indicator weights using the Group Analytic Hierarchy Process (Group-AHP), incorporating expert opinions from multiple fields to enhance objectivity. Indicator scores are quantified through three approaches: using standards and regulations, applying a pre-trained BERT model for text mining of inspection documents, and on-site evaluation by supervisory personnel. Finally, the safety risk level of boiler operation is assessed by integrating the indicator scores with their corresponding weights. The proposed framework leverages artificial intelligence techniques, stream data processing, and event-driven mechanisms to enable real-time risk identification, dynamic state modeling, and early hazard prediction. The evaluation outputs can offer valuable references for enterprises, inspection bodies, and regulatory authorities.