基于机器学习的转炉冶炼终点残锰含量预测  

Prediction of residual manganese content at the end point of converter smelting based on machine learning

作　　者：张龙强 闵义[1,2] 刘承军 黄健[3] 郑传新 马威[3] ZHANG Longqiang;MIN Yi;LIU Chengjun;HUANG Jian;ZHENG Chuanxin;MA Wei(Key Laboratory for Ecological Metallurgy of Multimetallic Ministry(Ministry of Education),Shenyang 110819,China;School of Metallurgy,Northeastern University,Shenyang 110819,China;Benxi Iron and Steel(Group)Co.,Ltd.,Benxi 117000,China)

机构地区：[1]多金属共生矿生态化冶金教育部重点实验室,辽宁沈阳110819 [2]东北大学冶金学院,辽宁沈阳110819 [3]本溪钢铁集团有限公司,辽宁本溪117000

出　　处：《炼钢》2024年第5期38-43,共6页Steelmaking

基　　金：国家自然科学基金面上资助项目(51974075)。

摘　　要：钢水锰含量控制主要在转炉炉后脱氧合金化阶段实现,转炉终点锰含量的准确获取对于锰合金加入量的确定具有重要影响,进而影响到钢水锰含量的精确控制。利用某钢厂转炉1500炉次历史冶炼生产数据,采用支持向量回归算法(SVR)、轻量级梯度提升机算法(LGBM)、分类梯度提升算法(CatBoost)对转炉终点残锰含量进行了预测,再使用贝叶斯优化算法(BayesSearchCV,BOA)分别对其优化。结果表明,贝叶斯优化后的CatBoost算法(BOA-CatBoost)效果最好,其决定系数R^(2)、均方误差MSE和均方根误差RMSE分别可达到0.712,0.000048和0.007021。残锰质量分数真实值与预测值的误差在±0.010、±0.008范围内,残锰含量预测命中率分别可达到83.2%和76.2%。The control of manganese content in molten steel is mainly completed during deoxidation and alloying process after oxygen blowing.The accurate acquisition of manganese content at the end point of the converter has an important influence on the determination of the addition amount of manganese alloy,and then affects the precise control of manganese content in molten steel.The residual manganese content at the end of the converter was predict via support vector regression(SVR),light gradient boosting machine(LGBM)and categorical boosting(CatBoost)algorithms based on the historical production data of 1500 heats in a steel works,and the prediction accuracy was improved via BayesSearchCV method.The results show that the optimized CatBoost algorithm(BOA-CatBoost)has the most advantage on the prediction,and its coefficient of determination R^(2),mean square error MSE and root mean square error RMSE can reach 0.712,0.000048 and 0.007021,respectively.Within the range of±0.010 and±0.008 of the errors between the real and predicted values of the residual manganese mass fraction,the prediction accuracy of residual manganese content can reach 83.2%and 76.2%,respectively.

关 键 词：转炉终点 残锰含量 机器学习 预测 

分 类 号：TF713.6[冶金工程—钢铁冶金]