利用改进XGBoost预测RH精炼终点钢水温度  

Predicting the endpoint steel temperature ofRH refining using improved XGBoost

作　　者：徐猛 雷洪[1,2] 何江一 韩信 王兴华 王燕斌 Xu Meng;Lei Hong;He Jiangyi;Han Xin;Wang Xinghua;Wang Yanbin(Key Laboratory of Electromagnetic Processing of Materials,Ministry of Education,Northeastern University,Shenyang 110819,China;School of Metallurgy,Northeastern University,Shenyang 110819,China;Jianlong Beiman Special Steel CoLtd,Qiqihar 161041,China;Shenyang Donghuike International Trade Company,Limited.Shenyang 113000,China)

机构地区：[1]东北大学材料电磁过程研究教育部重点实验室,沈阳110819 [2]东北大学冶金学院,沈阳110819 [3]建龙北满特殊钢有限责任公司,黑龙江齐齐哈尔161041 [4]沈阳东惠科国际贸易有限公司,沈阳113000

出　　处：《材料与冶金学报》2023年第5期437-443,共7页Journal of Materials and Metallurgy

基　　金：国家自然科学基金与宝钢联合资助项目(U1460108)。

摘　　要：合适的RH精炼终点钢水温度是保证连铸坯质量的重要前提.为了准确地预测RH精炼终点钢水温度,首先利用随机森林在处理高维数据集上的优势,采用袋外数据评分法进行特征选择,以确定模型的输入变量;然后利用Optuna框架优化XGBoost超参数,减少计算耗时,提高模型命中率;最后,结合实际的RH精炼生产数据进行模型验证.结果表明:与网格搜索和随机搜索相比,Optuna框架在优化XGBoost超参数的计算耗时和命中率上具有一定优势;与优化后的GBDT和LightGBM两种模型相比,优化后的XGBoost模型具有更好的预测性能,在偏差为±5℃时RH精炼终点钢水温度的命中率达到92%.The appropriate temperature of molten steel at the end of RH refining is an important prerequisite for ensuring the quality of continuous casting slabs.To predict the temperature of molten steel at the end of RH refining accurately,the out-of-bag scoring method is applied to determine the input variable of the model by character selection based on the advantages of random forest processing a multi-dimensional dataset at first.Secondly,the Optuna framework is applied to optimize XGBoost hyper-parameters to improve the hitting accuracy and the calculation time of XGBoost.Finally,such a system is applied to treat the industrial data about RH refining production.Compared with grid search and random search,the hitting accuracy is higher and the computing time is less if the Optuna framework is applied to optimize XGBoost hyper-parameters.Compared with the optimized GBDT model and the optimized LightGBM model,the optimized XGBoost model has better prediction performance,and the hit rate for the temperatures of molten steel±5℃is up to 92%at the end of RH vacuum refining.

关 键 词：RH精炼 XGBoost算法 终点预报 Optuna 钢水温度 

分 类 号：TF769.4[冶金工程—钢铁冶金]