机器学习在轻质高熵合金设计中的应用  

The application of machine learning in the design of lightweight high-entropy alloys

作　　者：陈巧冰 赫梓建 刘璇 赵艺 王殿辉[1,2] 胡朝浩 王仲民[2] CHEN Qiaobing;HE Zijian;LIU Xuan;ZHAO Yi;WANG Dianhui;HU Chaohao;WANG Zhongmin(School of Materials Science and Engineering,Guilin University of Electronic Technology,Guilin 541004,China;Guangxi Academy of Science,Nanning 530007,China)

机构地区：[1]桂林电子科技大学材料科学与工程学院,广西桂林541004 [2]广西科学院,南宁530007

出　　处：《桂林电子科技大学学报》2024年第3期237-245,共9页Journal of Guilin University of Electronic Technology

基　　金：国家自然科学基金(52061006);广西自然科学基金(2020GXNSFAA159122);广西信息材料重点实验室基金(131022-Z);广西碳酸钙资源综合利用重点实验室开放项目(HZXYKFKT202303,HZXYKFKT202206);广西科技基地与人才专项(桂科AD19110120)。

摘　　要：轻质高熵合金是一类相较于传统的合金具有更低密度和更高强度的新型材料,在航空航天、汽车工业、电子设备等高新产业有广泛的应用前景。然而,现阶段报道的高熵合金的强度与塑性的协调是亟待解决的一大难题,且制备工艺不完善,生产成本较高,暂时难以实现大规模推广与应用。因此,具有低密度和高强度的新型轻质高熵合金的设计与开发备受关注。着眼于准确、高效的机器学习方法在材料设计领域的应用与发展,基于国内外学者的研究成果,对利用机器学习低成本加速高熵合金设计进行了阐述,包括采用高通量实验与主动学习等方法不断挖掘、清洗、迭代、优化数据以及利用特征工程提高模型预测精度和性能,利用算法探寻标签值与特征变量间的关系,深入分析高熵合金性能强化的原因,并对近期机器利用学习预测轻质高熵合金的相结构及力学性能的研究进展进行了总结和展望。Lightweight high-entropy alloys(LHEAs)are a new class of materials that exhibit lower density and superior specific strength compared to traditional alloys.They hold extensive potential for applications in advanced industries such as aerospace,automotive,and electronics.However,the current challenge lies in balancing the strength and ductility of LHEAs,as well as improving the manufacturing processes to reduce production costs and facilitate large-scale adoption.Consequently,the design and development of new LHEAs with low density and high strength have garnered significant attention.The precise and efficient application of machine learning methods in material design is observed by leveraging research findings from scholars both domestically and internationally.It elaborates on the use of machine learning to expedite the design of LHEAs at a low cost,including the implementation of high-throughput experiments and active learning strategies to continuously mine,clean,iterate,and optimize data.Additionally,it discusses the enhancement of model prediction accuracy and performance through feature engineering,the exploration of relationships between label values and feature variables using algorithms,and a deep analysis of the underlying reasons for the performance enhancement of LHEAs.The paper concludes with a summary and outlook on recent advancements in predicting the phase structure and mechanical properties of lightweight LHEAs using machine learning.

关 键 词：机器学习 轻质高熵合金 相结构 力学性能 

分 类 号：TP181[自动化与计算机技术—控制理论与控制工程] TB301[自动化与计算机技术—控制科学与工程]