计算机应用基于机器学习模型的中锰钢临界温度A1和A3的预测  

Prediction of critical temperature A_(1) and A_(3) of medium-Mn steel based on machine learning models

作　　者：张志业 王焱 张彪 季泽 刘亚良 张明赫 冯运莉 Zhang Zhiye;Wang Yan;Zhang Biao;Ji Ze;Liu Yaliang;Zhang Minghe;Feng Yunli(College of Metallurgy and Energy,North China University of Science and Technology,Tangshan Hebei 063210,China;Shijiazhuang Customs Technology Center Caofeidian Business Department,Tangshan Hebei 063205,China)

机构地区：[1]华北理工大学冶金与能源学院,河北唐山063210 [2]石家庄海关技术中心曹妃甸业务部,河北唐山063205

出　　处：《金属热处理》2025年第2期268-277,共10页Heat Treatment of Metals

基　　金：国家自然科学基金(51901078);河北省自然科学基金(E2022209070);河北省中央引导地方科技发展资金(236Z1003G);唐山市市级科技计划(24130207C)。

摘　　要：为便于中锰钢热处理工艺的设计,开发了用于中锰钢临界温度A_(1)、A_(3)预测的机器学习模型。通过Thermal-Calc模拟软件获取496组不同成分中锰钢临界温度数据,以Mn、Al、C成分作为输入特征,以相变温度A_(1)和A_(3)作为输出目标。采用均方根误差(RMSE)、平均绝对误差(MAE)和决定系数(R^(2))3种指标对模型预测效果进行评价。从7种机器学习模型(LR、DT、SVM、GPR、Boosting、Bagging以及ANN)中筛选出了预测A_(1)的GPR模型和A_(3)的GPR、ANN模型。结果表明,预测A_(1)的GPR模型具有足够精度,为预测A_(1)的最优模型。采用网格搜索法对预测A_(3)的初步模型进行超参数调节,从而获得A_(3)的最优模型(单层ANN模型)。针对所应用文献中中锰钢的化学成分,利用最优模型对A_(1)、A_(3)进行预测,得到A_(1)、A_(3)预测值与实测值的总体MAE分别为9.95℃和13.57℃,最低差距分别为0.30℃和6.20℃,表明模型精准度高,可用于中锰钢临界温度的预测。In order to facilitate the design of heat treatment process of medium-Mn steel,a machine learning model for predicting the critical temperature A_(1)and A_(3)of medium-Mn steel was optimized.The critical temperature data of 496 groups of medium-Mn steels with different compositions were obtained by Thermal-Calc simulation software.Mn,Al and C compositions were taken as input characteristics,and phase transition temperatures A_(1)and A_(3)were taken as output targets.Three indexes of root mean square error(RMSE),mean absolute error(MAE)and determination coefficient(R^(2))were used to evaluate the prediction effect of the model.From seven machine learning models(LR,DT,SVM,GPR,Boosting,Bagging and ANN),the GPR model for predicting A_(1)and the GPR and ANN model for predicting A_(3)were screened.The results show that the GPR model for predicting A_(1)has sufficient accuracy,that is the optimal model for A_(1).The grid search method is used to adjust the hyperparameters of the preliminary model for predicting A_(3),and the optimal model of A_(3)(single-layer ANN model)is obtained.According to the chemical composition of medium-Mn steel in the applied literature,A_(1)and A_(3)are predicted by using the optimal model.The overall MAE of the predicted value and measured value of A_(1)and A_(3)is 9.95℃and 13.57℃,respectively,and the minimum difference is 0.30℃and 6.20℃,respectively,indicating that the model has high accuracy and can be used to predict the critical temperature of medium-Mn steel.

关 键 词：机器学习 中锰钢 临界温度 参数调节 

分 类 号：TG111.7[金属学及工艺—物理冶金]