机器学习方法在高分子玻璃化研究中的应用  

Applications of Machine Learning Methods in the Studies of Polymer Glass Formation

作　　者：杨镇岳 聂文建 刘伦洋 徐晓雷[1] 夏文杰 徐文生 Zhen-yue Yang;Wen-jian Nie;Lun-yang Liu;Xiao-lei Xu;Wen-jie Xia;Wen-sheng Xu(State Key Laboratory of Polymer Physics and Chemistry,Changchun Institute of Applied Chemistry,Chinese Academy of Sciences,Changchun 130022;Department of Civil,Construction and Environmental Engineering,North Dakota State University,Fargo,North Dakota 58108)

机构地区：[1]中国科学院长春应用化学研究所、高分子物理与化学国家重点实验室,长春130022 [2]Department of Civil,Construction and Environmental Engineering,North Dakota State University,Fargo,North Dakota 58108

出　　处：《高分子学报》2023年第4期432-450,共19页Acta Polymerica Sinica

基　　金：国家自然科学基金(基金号22222307,21973089,22203087);北达科他州立大学工程学院资助项目。

摘　　要：高分子玻璃的物理性质与其结构和动力学密切相关.揭示高分子玻璃化的微观物理图像对高分子玻璃材料的结构调控和分子设计至关重要.然而,高分子的长链结构和复杂单体结构特征致使目前仍然缺乏普适的理论或者模型来定量解释高分子玻璃化的物理机制.因此,亟需发展更为先进的研究方法从而更深入地理解高分子玻璃化.近年来,国内外学者利用基于数据驱动的信息学方法(例如机器学习)对高分子玻璃化开展了研究,并取得了丰富成果.本综述首先介绍了常用的高分子信息学数据库和机器学习算法.之后,从高分子玻璃化转变温度的预测、新型高分子玻璃材料的研发、过冷液体的结构-动力学关系和玻璃体系相变的确定四个方面总结和评述了机器学习应用在玻璃化研究中的代表性进展.最后,探讨了机器学习方法在高分子玻璃化研究中面临的主要挑战,并对玻璃信息学这一领域的发展进行了展望.The physical properties of polymer glasses are closely related to their structure and dynamics.A deep understanding of the microscopic physical mechanism of polymer glass formation is crucial for the structural control and molecular design of polymer glasses.However,various molecular characteristics associated with complex segmental structures and chain topology of polymers impose significant challenges for the development of a fully predictive theory to describe their glass formation in a quantitative way.Hence,it is highly desired to develop more advanced approaches to better understand and predict polymer glass formation.In recent years,growing efforts have been made to shed new lights on polymer glass formation based on the data-driven informatics approaches,such as machine learning,and important progresses have been made towards this direction.The present review first introduces common polymer databases and machine learning algorithms,followed by a summary and review of the representative progresses on the applications of machine learning methods in the studies of polymer glass formation.In particular,the focus is placed on the prediction of the glass transition temperature,the research and development of novel glassy polymer materials,the investigation of structure-dynamics relationships of glass-forming liquids,and the determination of phase transitions of glasses.Finally,the present review discusses challenges and opportunities in the applications of machine learning methods to polymer glass formation and provides a perspective on glass informatics.

关 键 词：信息学 高分子玻璃化 机器学习 理论模拟 

分 类 号：TP181[自动化与计算机技术—控制理论与控制工程] TQ317[自动化与计算机技术—控制科学与工程]