含能材料2024，Vol.32Issue(6)：573-583,11.DOI:10.11943/CJEM2024055
机器学习辅助的[5,6]稠环含能化合物高通量设计
Machine Learning Assisted High-throughput Design of[5,6]Fused Ring Energetic Compounds
摘要
Abstract
Compared with the research and development model guided by experience and calculations,machine learning-assisted high-throughput virtual screening technology for energetic molecules has shown obvious advantages in terms of molecular design efficiency and quantitative analysis of structure-activity relationships.In view of the fact that nitrogen-rich fused ring energetic compounds usually show better energy-stable balance properties,this study uses machine learning-assisted high-throughput virtual technology to conduct chemical space exploration of[5,6]nitrogen-rich fused ring energetic molecules.Based on the[5,6]all-carbon skeleton,this study obtained 142,689[5,6]fused ring compounds through combined enumera-tion and aromatic screening.At the same time,a machine learning algorithm was used to establish and optimize an energetic molecular property prediction model(including density,decomposition temperature,detonation velocity,detonation pressure,impact sensitivity and enthalpy of formation).The effects of nitrogen and oxygen atoms on the fused ring and functional groups on the molecule on the performance of energetic compounds were analyzed.The research results show that the structure-activity relationship of the generated fused ring compounds is consistent with the general correlation between energy and stability of ener-getic compounds,verifying the rationality of the prediction model.Taking detonation velocity and decomposition temperature as the criteria for energy and thermal stability,five molecules with outstanding comprehensive properties were screened,and the quantum chemical calculation results were in good agreement with the machine learning prediction results,which further veri-fied the accuracy of the prediction model.关键词
机器学习/高通量筛选/核岭回归/分子设计/[5,6]稠环含能化合物Key words
machine learning/high-throughput screening/kernel ridge regression/molecular design/[5,6]fused ring energetic compounds分类
军事科技引用本文复制引用
潘林虎,王睿辉,樊明仁,宋思维,王毅,张庆华..机器学习辅助的[5,6]稠环含能化合物高通量设计[J].含能材料,2024,32(6):573-583,11.基金项目
国家自然科学基金(22205218,22075259,22175157) National Natural Science Foundation of China(Nos.22075259,22175157,22205218) （22205218,22075259,22175157）
