摘要
钢的回火力学性能曲线是回火工艺参数优选的基础,机器学习为回火工艺参数的优选提供了新的高效途径。本工作采用ANN、IBK、Bagging、Randomtree和M5Rules等多种机器学习模型算法预测合金结构钢的回火力学性能。2950组钢回火性能数据提取于国家材料科学数据共享网,以化学成分、回火温度作为输入特征,以抗拉强度、屈服强度和维氏硬度作为输出目标。采用相关系数(R)、均方根误差(RMSE)、平均绝对误差(MAE)和相对误差(δ)进行模型的评估与定型。结果表明:回火抗拉强度、屈服强度、维氏硬度的预测分别采用IBK、Randomtree和Bagging算法能够得到更高的预测精度,相对误差分别集中于±6%、±10%、±10%。使用最佳模型对测试集四种钢的回火力学性能预测结果良好,预测精度高于JMatPro软件和经验公式的计算结果。限于已有成分特征数据分布的不均衡,为进一步扩展模型的预测精度与泛化性,需积累更多关于微合金V、Ti、B等方面的数据。
The tempering mechanical property curve of steel is the basis for the optimization of tempering process parameters,and machine learning provides a new and efficient way for the optimization of tempering process parameters.In this paper,many kinds of machine learning model algorithms such as ANN,IBK,Bagging,Randomtree and M5Rules are used to predict the tempering mechanical properties of alloy structural steel.The tempering performance data of 2950 groups of steels are extracted from the National Materials Science Data Sharing Network,with chemical composition and tempering temperature as input characteristics,and tensile strength,yield strength and Vickers hardness as output targets.Correlation coefficient(R),root mean square error(RMSE),average absolute error(MAE)and relative error(δ)are used to evaluate and finalize the model.The results show that higher prediction accuracy can be obtained by using IBK,Randomtree and Bagging algorithms to predict tempering tensile strength,yield strength and Vickers hardness,and the relative errors are concentrated between±6%,±10%and±10%,respectively.The best model is used to predict the tempering mechanical properties of four kinds of steel in the test set,and the prediction accuracy is higher than that of JMatPro software and empirical formula.Limited to the unbalanced distribution of the existing composition characteristic data,in order to further expand the prediction accuracy and generalization of the model,it is necessary to accumulate more data about microalloy V,Ti,B and so on.
作者
高志玉
樊献金
高思达
薛维华
GAO Zhiyu;FAN Xianjin;GAO Sida;XUE Weihua(School of Materials Science and Engineering,Liaoning Technical University,Fuxin 123000,Liaoning,China;School of Materials Science and Engineering,Shenyang Ligong University,Shenyang 110159,China)
出处
《材料导报》
EI
CAS
CSCD
北大核心
2023年第6期144-150,共7页
Materials Reports
基金
辽宁省教育厅高等学校基本科研项目(LJ2020JCL021,LJKMZ20220593)。
关键词
机器学习
回火力学性能
合金结构钢
算法模型
machine learning
tempering mechanical property
alloy structural steel
algorithmic model
