基于机器学习的高速棒材力学性能预测  

Prediction of mechanical properties of high speed bar based on machine learning

作　　者：赵奕光 王卫卫[1] 张宏亮[1] 肖金福[1] 赵飞宇 冯光宏[1] Zhao Yiguang;Wang Weiwei;Zhang Hongliang;Xiao Jinfu;Zhao Feiyu;Feng Guanghong(Metallurgical Technology Institute of Central Iron&Steel Research Institute,Beijing 100081,China)

机构地区：[1]钢铁研究总院冶金工艺研究所,北京100081

出　　处：《金属制品》2025年第1期56-60,共5页Metal Products

摘　　要：高速棒材轧制相较板材具有头尾温差大、横截面组织均匀性差、应变速率快等特点,在实际生产过程中面临着大量重复的组织检验,复杂的工艺参数调控问题,增加生产成本的同时,生产效率低。针对上述问题,提出一种基于机器学习的高速棒材力学性能预测模型,采用BP神经网络、支持向量机、随机森林3种机器学习模型算法预测高速棒材力学性能。高速棒材以HRB500E为研究对象,数据集来源于实际生产大数据,以化学成分、轧制工艺等为输入特征,以抗拉强度、屈服强度和强屈比作为输出目标。采用相关系数R^(2)、均方根误差RMSE、平均绝对误差MAE进行模型的评估。结果表明支持向量机SVR模型算法具有更好的预测精度,相关系数分别达到0.96、0.98、0.923,可以应用于高速棒材在线性能预测模型中,作为监控、分析高速棒材力学性能及优化工艺的有效工具。Compared with plate rolling,high speed bar rolling has characteristics of large temperature difference between head and tail,poor cross-sectional microstructure uniformity,and fast strain rate.In actual production process,a large number of microstructure tests and complex process parameter control increase production cost,resulting in low production efficiency.To solve above problems,prediction model of mechanical properties of high speed bar based on machine learning is proposed.BP neural network,support vector regression and random forest are used to predict mechanical properties of high speed bar.HRB500E is research object of high speed bar.Data set is derived from actual production big data.Chemical composition and rolling process are input characteristics,and tensile strength,yield strength and strength yield ratio are output targets.Correlation coefficient R^(2),root mean square error RMSE and mean absolute error MAE were used to evaluate model.Results show that support vector machine SVR model algorithm has better prediction accuracy,and correlation coefficients are 0.96,0.98 and 0.923 respectively.It can be applied to online performance prediction model of high speed bar,and can be used as an effective tool for monitoring,analyzing mechanical properties of high speed bar and optimizing process.

关 键 词：HRB500E 高速棒材 机器学习 力学性能 支持向量机 

分 类 号：TG335.62[金属学及工艺—金属压力加工]