摘要
目的:找到内禀矫顽力(H_(cj))、剩磁(B_(r))、最大磁能积((BH)_(max))和配方成分、工艺参数之间的内在关联性。方法:我们从相关论文收集了140余条数据,使用配方成分、工艺参数作为数据特征,使用内禀矫顽力、剩磁、最大磁能积作为目标属性,采用随机森林算法和梯度提升树的机器学习算法构建预测模型,最后通过我们实验室的数据来测试模型的预测能力。结果:3个模型的平均绝对误差(MAE)分别为91.8 kA/m、0.047 T和6.583 kJ/m^(3)。H_(cj)的预测值与实验室测得数据的MAE为78.0 kA/m。预测数据与实际数据基本一致。结论:使用梯度提升树构建的预测模型在配方成分和工艺参数对磁性能的映射上有良好的预测性,可以为寻找低成本高性能的磁体提供指导,有利于从根本上提高混合稀土永磁合金的研究速度。
Aims:This paper aims to find the intrinsic correlation between intrinsic coercivity(H_(cj)),remanent magnetization(B_(r)),maximum magnetic energy product((BH)_(max))and chemical compositions as well as process parameters.Methods:We collected more than 140 data from related papers,used chemical compositions and process parameters as data features,used intrinsic coercivity,remanence,and maximum magnetic energy product as target attributes,and used machine learning algorithms of random forest and gradient boosting tree to construct prediction models,and finally tested the prediction ability of the models with data from our laboratory.Results:The mean absolute errors(MAE)of the three models were 91.8 kA/m,0.047 T,and 6.583 kJ/m^(3),respectively.The predicted value of H_(cj)was 78.0 kA/m with the MAE of the measured data in the laboratory.The predicted data were basically consistent with the actual data.Conclusions:The prediction model constructed using the gradient boosting tree has good predictive properties in mapping the chemical composition and process parameters on the magnetic properties,which can provide guidance for the study of low-cost and high-performance magnets.
作者
谢昊
席子昱
吴琼
XIE Hao;XI Ziyu;WU Qiong(College of Materials and Chemistry,China Jiliang University,Hangzhou 310018,China)
出处
《中国计量大学学报》
2022年第1期146-150,共5页
Journal of China University of Metrology
基金
国家重点研发计划子课题(No.2019YFF021705)
中国计量大学基本科研业务费项目(No.2020YW24,2020YW32)。
