人工智能辅助含能分子设计的应用与展望  

Applications and Prospects of AI⁃assisted Design of Energetic Molecules

作　　者：刘锐 刘建[1] 唐岳川 张朝阳[1] 黄静 黄鑫[1] LIU Rui;LIU Jian;TANG Yue-chuan;ZHANG Chao-yang;HUANG Jing;HUANG Xin(Institute of Chemical Materials,CAEP,Mianyang 621999,China;School of Chemistry and Chemical Engineering,Southwest Petroleum University,Chengdu 610500,China)

机构地区：[1]中国工程物理研究院化工材料研究所,四川绵阳621999 [2]西南石油大学化学化工学院,四川成都610500

出　　处：《含能材料》2024年第4期408-421,共14页Chinese Journal of Energetic Materials

基　　金：国家自然科学基金(22173086,22203081,22305234)。

摘　　要：含能分子研发面临多重挑战,传统“试错法”效率低下,计算机辅助分子设计的出现改变了研发模式。本综述回顾了含能分子设计的发展历程,介绍了计算机辅助含能分子设计的研究现状,并概述了人工智能技术(AI)在性质预测、分子生成、合成路线和反应条件预测等多个设计环节的最新进展,讨论了当前含能分子设计模式与其他材料设计方法的差距,思考差距产生的原因,并对未来AI辅助含能分子设计的发展方向提出展望。研究发现,AI在含能分子性能预测和分子生成等方面已经有了应用,但在合成路径规划和反应条件优化等环节的应用仍有待进一步探索,应用前景巨大。通过数据增强、迁移学习或高通量计算有望能够解决含能分子数据薄弱的问题;加强AI辅助含能分子合成路线与反应条件探索有望贯通“设计→评估→制备→验证”全流程自动化分子设计模式。AI辅助含能分子设计为提升含能分子设计水平提供新的可能性,有助提升含能分子研发效率。The explore of energetic molecules faces multiple challenges,and the traditional design method are inefficient.The emergence of computer‑aided molecular design has changed the research and development model.This review provides an overview of the development of energetic molecular design and introduces the current research status of computer‑aided energetic molecular design.By summarizing the latest advancements in Artificial Intelligence(AI)technology across various design aspects,including performance prediction,molecular generation,retrosynthetic reaction prediction,and reaction condition prediction,we discussed the existing gap between the current approaches in energetic molecular design and other materials design methods.By thinking about the causes of the gap,we present an outlook on the future developmental directions of AI‑assisted energetic molecular design.Research indicates that AI has already been applied in property prediction and molecular generation of energetic molecular design,but requires further exploration in retrosynthetic reaction prediction,and reaction conditions prediction.AI‑assisted design of energetic molecules holds broad promising application prospects.Data enhancement,transfer learning and high‑throughput computing are expected to solve the problem of weak data of energetic molecules.Enhancing AI‑assisted prediction of synthesis routes and reaction conditions for energetic molecules shows promise for achieving the automatic molecu‑lar design via whole process of“design→evaluation→preparation→verification”.AI‑assisted energetic molecular design provides new possibilities for improving the level of energetic molecular design and helps to improve the efficiency of energetic molecule research and development.

关 键 词：含能分子 分子设计 人工智能 机器学习 定量构效关系 

分 类 号：TJ55[兵器科学与技术—军事化学与烟火技术] O62[理学—有机化学]