Quantitative risk assessment of explosion rescue by integrating CFD modeling with GRNN

Gas explosions remain a significant industrial hazard, characterized by sudden outbreak, rapid development, and huge destruction. Quantitative risk assessment (QRA) has played an effective role in safety management and emergency preparedness for such incidents. Although a lot of attempts have been done to analyze the explosion risks, few works have been conducted on the risk for responding rescuers during these missions. This paper presents an explosion rescue risk assessment methodology for emergency decision support by integrating with a general regression neural network (GRNN) with computational fluid dynamics (CFD) modeling. Underground coal mine gas explosions are taken as an example. The likelihood exposure consequence (LEC) method is combined with a fault tree to establish a rescue risk assessment model that consists of 5 levels. The CFD modeling for possible explosion scenarios is continuously performed by automatically varying the predefined parameters. The generated data is used for the development and improvement of the GRNN model. Provided with real-time data, the GRNN model can predict the effects of a blast within few seconds, which are then used to calculate the occurrence probability of secondary explosions. As a result, the rescuers’ level exposure to explosion risk can be estimated. This will allow a better-informed rescue decision making. The proposed integrated method is applied to the Laoyingyan coal mine to demonstrate its applicability and effectiveness.