Prediction of Initial Reaction Characteristics of Materials from Molecular Conformational Changes Based on Artificial Intelligence Technology

To determine microscopic reaction mechanisms of energetic materials, a problem exists when there are multiple calculations but limited calculation scales. Herein, we used artificial intelligence algorithms of a convolutional neural network and a multilayer perceptron to establish a prediction model. This model was based on the storage and conversion mechanisms of shock energy and molecular conformational change as well as the reaction mechanism obtained using molecular dynamics simulation. Further, based on the changes in conformational parameters, such as bond length, bond angle, and dihedral angle, the molecular volume change degree was predicted and then the initial bond breaking and product generation probabilities were predicted according to the molecular volume change degree. Consequently, when the molecules were loaded with shock energy, we could realize the rapid assessment of molecular conformational changes and reaction processes. The accuracy and universality of the prediction model were verified by the agreement between the prediction results of the mechanism quantification models and the reactive molecular dynamics simulation results of multiple energetic materials. Our artificial intelligence prediction method can predict the energy storage and conversion mechanisms as well as material conformational transformation and reaction properties of materials with a smaller computational load and higher computational analysis efficiency than molecular dynamics simulation and analysis.