基于机器学习的热轧含Ti高强钢性能预测  

Performance prediction of Ti-containing high-strength steel in hot rolling based on machine learning

作　　者：赵江涛 周正军 刘斌 邹航 ZHAO Jiangtao;ZHOU Zhengjun;LIU Bin;ZOU Hang(Technology Center,Wuhan Iron&Steel Co.,Ltd.,Wuhan 430083,Hubei,China;Equipment Management Department,Wuhan Iron&Steel Co.,Ltd.,Wuhan 430083,Hubei,China)

机构地区：[1]武汉钢铁有限公司技术中心,湖北武汉430083 [2]武汉钢铁有限公司设备管理部,湖北武汉430083

出　　处：《宝钢技术》2025年第1期50-56,共7页Baosteel Technology

摘　　要：通过材料成分和工艺参数对钢的性能进行预测可有效降低生产成本、提高产品质量,因此该方法受到广泛关注。针对含Ti高强钢在生产过程中体现出的性能波动大、生产不稳定问题,基于简单循环单元(simple recurrent unit,SRU)模型,根据真实生产数据,使用机器学习模型预测含Ti高强钢的屈服强度、抗拉强度和延伸率。结果显示SRU模型具有较高的精度:抗拉强度预测值和实际值相对误差在±5%以内的命中率为88.63%,相对误差在±10%以内命中率达到97.80%;屈服强度预测值和实际值相对误差在±5%以内的命中率为94.17%,相对误差在±10%以内的命中率达到99.52%;延伸率预测值和实际值相对误差在±5%以内的命中率为74.50%,相对误差在±10%以内的命中率达到95.15%。该模型可以有效地降低含Ti高强钢生产成本,缩短开发周期。该研究为热轧带钢产品力学性能预测等需要考虑多维度及时间序列的问题提供了新的方法。Predicting the performance of steel using chemical composition and process parameters can effectively reduce production costs and improve product quality,thus garnering widespread attention.In response to the issue of significant performance fluctuations and production instability exhibited during the manufacturing process of Ti-containing high-strength steel,this study employs a simple recurrent unit(SRU)model to predict the yield strength,tensile strength,and elongation of Ti-containing high-strength steel based on actual production data using machine learning techniques.The results demonstrate that the SRU model achieves sufficiently high accuracy:the hit rate for the relative error between the predicted and actual tensile strength within±5%is 88.63%,and within±10%it reaches 97.80%;the hit rate for the relative error between the predicted and actual yield strength within±5%is 94.17%,and within±10%it reaches 99.52%;the hit rate for the relative error between the predicted and actual elongation within±5%is 74.50%,and within±10%it reaches 95.15%.The model can effectively reduce the production costs and shorten the development cycles of Ti-containing high-strength steel.This study provides a new method for predicting the mechanical properties of hot-rolled strip steel products and other issues that need to consider multiple dimensions and time series.

关 键 词：含Ti钢 性能预测 机器学习 屈服强度 

分 类 号：TG113.25[金属学及工艺—物理冶金]