Prediction of Explosion Shock Wave Parameters of CHON-Type Explosive using Machine Learning Model

The accidental explosion of explosives poses significant threats to lives and property. Accurately predicting the shock wave parameters and damage effects of explosions has become a subject of widespread interest. In this study, a CHON explosive is selected as the research object, physics-informed machine learning-aided framework (PIMLAF) and artificial neural network (ANN) are proposed to predict the shock wave parameters of the CHON explosive. By using scaled distance and detonation heat as input parameters, the framework predicts the peak overpressure, shock wave arrival time, positive pressure duration, shock wave attenuation coefficient, and the overpressure-time curve at different scaled distances. Compared to the ANN, PIMLAF improves prediction accuracy by 27.4% to 60.0% under the condition of uneven data distribution. Compared with physical models, the percentage of predictions with an error margin of less than 10% increases from 27.3%–79.5% to 82.1%–100% using PIMLAF. Furthermore, compared to finite element models, PIMLAF enhances computational efficiency by tens to hundreds of times while maintaining a calculation accuracy of no less than 5% and significantly reducing the required input parameters. The proposed PIMLAF is simpler and more efficient, offering a robust tool for predicting explosive shock wave parameters. The research findings provide valuable references for evaluating explosive power fields and improving explosion safety measures.