Dynamic intelligent risk assessment of hazardous chemical warehouse fire based on electrostatic discharge method and improved support vector machine

Abstract

Chemical accidents occur frequently in China owing to inadequate process safety risk management and control in warehouses. Real-time dynamic risk assessment can identify stored process risks and reduce the accident probability. The support vector machine (SVM) is an effective dynamic risk assessment method. To improve the dynamic risk assessment performance of the SVM model, the electrostatic discharge method (ESDA), which has a strong optimization ability, was used to optimize the model parameters. An improved mixed kernel (NP mixed kernel) that was a linear combination of the novel radial basis function and polynomial kernel was constructed, and an intelligent assessment model of the warehouse fire dynamic risk based on the ESDA and improved SVM (ESDA-NPSVM) was proposed. The experimental results indicated that the proposed model had excellent performance for the dynamic risk assessment of fire accidents in Class A hazardous chemical warehouses, suggesting that it is useful for practical applications.