The accurate prediction of thermodynamic properties of energetic materials is crucial for advancing the development of safer and more efficient explosives, propellants, and pyrotechnics. This research presents a comprehensive machine learning approach to predict the enthalpy of formation of energetic materials using molecular descriptors and functional group information. By analyzing a dataset of 113 distinct energetic compounds, we developed and compared multiple machine learning models, including Decision Trees, Random Forest, AdaBoost, K-Nearest Neighbors, Ensemble Learning, Multilayer Perceptron Artificial Neural Networks, and Convolutional Neural Networks. The CNN model demonstrated superior generalization capabilities with an R2 value of 0.83 for the total dataset and minimal overfitting between training and testing phases. Feature importance analysis using SHAP revealed that azide functional groups have the most significant influence on enthalpy prediction, aligning with established thermochemical principles. This predictive framework offers a reliable computational alternative to expensive experimental methods for screening novel energetic materials.
Li et al. (Sun,) studied this question.