机器学习在PUREX工艺中的应用综述  

A review on the application of machine learning in PUREX

作　　者：于婷[1] 张音音 金文蕾 张睿志 罗应婷 朱升峰 何辉[1] 龚禾林 叶国安[1] YU Ting;ZHANG Yin-Yin;JIN Wen-Lei;ZHANG Rui-Zhi;LUO Ying-Ting;ZHU Sheng-Feng;HE Hui;GONG He-Lin;YE Guo-An(China Institute of Atomic Energy,Beijing 102413,China;School of Mathematics,Sichuan University,Chengdu 610064,China;School of Mathematical Sciences,East China Normal University,Shanghai 200241,China;Paris Elite Institute of Technology,Shanghai Jiao Tong University,Shanghai 200240,China)

机构地区：[1]中国原子能科学研究院,北京102413 [2]四川大学数学学院,成都610064 [3]华东师范大学数学科学学院,上海200241 [4]上海交通大学巴黎卓越工程师学院,上海200240

出　　处：《四川大学学报(自然科学版)》2025年第2期297-308,共12页Journal of Sichuan University(Natural Science Edition)

基　　金：国家自然科学基金(U2241289);中核集团青年英才基金(FY222506000503)。

摘　　要：为了控制核燃料的环境危害,必须对其进行回收管理和后处理.PUREX是一种高效的核燃料后处理方法,具有效率高、可扩展性强及适用性广等优点,目前已经得到了广泛应用.PUREX工艺很复杂,操作时必须对萃取剂及溶液的性质进行准确全面的测量和分析,并对处理流程进行严格保障.为了获取相关数据,已有研究大多依赖传统的批量实验测试法.这种方法需要多次实验迭代和人工操作,研究周期长、资源消耗大,且萃取效率低.机器学习主要通过学习和分析海量数据来实现模型预测,有望避免批量实验测试法中出现的问题,提高萃取效率.近年来,各种机器学习方法在PUREX中的应用逐渐增多.本文旨在对机器学习方法在PUREX流程的内部反应和后处理保障等方面的应用进行综述,促进和推动机器学习方法在PUREX中的进一步应用,提高萃取效率.在内部反应方面,本文概述了一些机器学习算法在萃取剂选择、萃取剂性质预测及溶液选择等方面的应用.在后处理保障方面,本文概述了一些机器学习算法在保障后处理流程顺利进行等方面的应用.综合看,将机器学习方法应用于PUREX流程确实能提供更精准的预测结果、提高萃取效率.此外,本文还对机器学习方法在PUREX相关领域的应用进展进行了概述,重点介绍了分子动力学与机器学习方法的集成方法及其应用,以便启发PUREX与机器学习模型相结合的方法的研究.最后,本文还介绍了一些利用机器学习方法扩展和增强数据集、克服数学模型建模时可用数据集过小问题的具体应用.With the development of nuclear power industry and the widespread applications of nuclear fuel in energy sector,the amount of nuclear waste generated increases rapidly.Due to the necessity of resource recovery and management of the radioactive materials to control hazards,nuclear fuel reprocessing is considered an indispensable process.Plutonium uranium refining by extraction(PUREX),owing to its high efficiency,strong scalability and wide applicability,has been extensively employed in nuclear fuel reprocessing.In PUREX,it is necessary to comprehensively measure and analyze the properties of extractants and solution systems and to ensure the safeguarding of PUREX reprocessing.Previous studies mainly focused on obtaining relevant data by traditional batch experiments involving multiple experimental iterations and manual operations,which results in long research cycle,increased resource consumption and low extraction efficiency.Machine learning(ML)is extensively expected to play a vital role in the prediction and decision-making of PUREX since ML can avoid the problem inherent in the repeated experiments by learning and analysis of vast amounts of data.In this paper,we focus on the application of ML in PUREX and provide a review on the application of ML in the internal reaction and reprocessing of PUREX process.On the one hand,the applications of ML in the selection of extractants,the prediction of extractants properties and the selection of solution systems are elucidated.On the other hand,the applications of ML for safeguarding the reprocessing process are introduced.It is positive that the applications of ML in PUREX can provide more accurate predictions and improve the extraction efficiency.We also delve into the applications of ML in other relevant fields of PUREX,such as the liquid-liquid extraction domain and nuclear engineering field,which lays the groundwork for the integrations of PUREX with other ML models.Specifically,we introduce a novel approach com‑bining molecular dynamics with ML which allows for

关 键 词：PUREX 机器学习 核燃料后处理 萃取剂 溶液体系 

分 类 号：O29[理学—应用数学]