Study on erosion prediction model of high-energy gun propellant based on machine learning

The erosion characteristics is an important reference index that affects the popularization and application of RDX-based high-energy gun propellant containing energetic plasticizer Diazidonitrazapentane (DIANP). For predicting the ablative properties of RDX-based high-energy gun propellants, the research prepared three kinds of propellant samples and performed erosion simulation tests. Based on the experimental results, the erosion kinetic model of the propellant gas to the barrel was derived from regression analysis and attribution analysis. Then, the Latin hypercube sampling method was used to generate a large number of the propellant formulations and working pressure combinations, and the ablative properties of these propellant formulations could be calculated by the erosion kinetic model. The propellant components and working pressure were set as characteristic values, and the corresponding ablative properties of these working conditions were taken as the response value to train the erosion prediction model of propellant based on machine learning. The results illustrated that the natural logarithm of mass loss and the reciprocal of explosion temperature showed strong linear relationships at different working pressure. Meanwhile, the influence of working pressure on erosion is more significant. Among the selected machine learning algorithms, the goodness of fit of the ensemble learning algorithm to sample data was 0.95, and the average relative deviation between the experimental data and predictive data was 8.93%, which demonstrated excellent prediction effect and generalization ability of the erosion prediction model based on ensemble learning. The aim of the erosion prediction model proposed in the study is to efficiently pre-evaluate the ablative property through the gun propellant formulation.