| 766 | 2 | 120 |
| 下载次数 | 被引频次 | 阅读次数 |
针对国内某钢厂的转炉终点控制模型受高炉铁水成分和温度波动较大等因素的影响，致使终点碳温预测命中率偏低的问题，本文中利用现场生产数据建立了基于机器学习的转炉终点智能控制模型，并使用不同的Boosting算法模型对转炉终点进行预测.结果表明：4种Boosting算法模型的预测准确率均高于机理模型的预测准确率，其中CatBoost模型的准确率最高，其预测值与真实值差距最小；在200炉次中，CatBoost模型终点钢水碳含量预测偏差在±0.02%以内的有166炉，命中率为83.0%,终点温度预测偏差在±15℃以内的有165炉，命中率为82.5%;与机理模型相比，终点钢水碳含量命中率提高了17个百分点，终点温度命中率提高了23.5个百分点，使用CatBoost模型预测能够为现场转炉冶炼过程终点判断提供有效指导.
Abstract：Aiming at the problem that the converter endpoint control model based on metallurgical mechanism and flue gas analysis in a domestic steel plant has a low hit rate of endpoint carbon temperature prediction due to the large fluctuation of molten iron composition and temperature in blast furnace, this study uses the field production data to establish an intelligent control model of converter endpoint based on machine learning, and uses different Boosting machine algorithm models to predict the converter endpoint. The results indicate that all four Boosting algorithm models outperformed the traditional metallurgical mechanism-based model in terms of prediction accuracy. Among these, the CatBoost model demonstrated the highest performance, with the smallest deviation between predicted and actual values. Of the 200 production batches, 166 batches exhibited a carbon content prediction error within 0.02%, achieving an accuracy rate of 83%, while 165 batches showed a temperature prediction error within 15 ℃, yielding an accuracy rate of 82.5%. Compared to the mechanism model, the CatBoost model improved the accuracy of endpoint carbon content prediction by 17 percentage points and temperature prediction by 23.5 percentage points. Therefore, the application of the CatBoost model can effectively support and guide the endpoint judgement of steelmaking process for in-site converters.
[1] 鲁川.转炉炼钢终点控制技术探讨[J].冶金与材料，2021(2):87-88.(Lu Chuan.Discussion on endpoint control technology of converter steelmaking[J].Metallurgy and Materials,2021(2):87-88.)
[2] 郭宝辉.用于转炉终点预测的光谱采集系统设计[D].南京：南京理工大学，2014.(Guo Baohui.Design of spectrum acquisition system for predicting the end point of converter [D].Nanjing:Nanjing University of Science and Technology,2014.)
[3] 潘玉柱，张彩军，朱新华，等.基于烟气分析的转炉连续碳温预报模型[J].河南冶金，2014,22(5):15-17.(Pan Yuzhu,Zhang Caijun,Zhu Xinhua,et al.Converter continuous carbon temperature prediction model based on flue gas analysis[J].Henan Metallurgy,2014,22(5):15-17.)
[4] Mitchell T M.Machine Learning[M].New York:McGraw-Hill,1997.
[5] Zhou Z H.Ensemble Methods:Foundations and Algorithms[M].Boca Raton:CRC Press/Taylor & Francis,2012.
[6] 高尉.Boosting方法的理论研究[D].南京：南京大学，2014.(Gao Wei.Theoretical recsearch on boosting[D].Nanjing:Nanjing University,2014.)
[7] Freund Y,Schapire R E.A decision-theoretic generalization of on-line learning and an application to boosting[J].Journal of Computer and System Sciences,1997,55(1):119-139.
[8] Freund Y,Schapire R E.Game throry,on-line prediction and boosting[C] // Proceedings of the 9th Annual Conference on Learning Theory.New York:Association for Computing Machinery,1996:325-332.
[9] Freund Y,Schapire R E.A decision-theoretic generalization of on-line learning and an application to boosting[J].Journal of Computer and System Sciences,1997,55(1):119-139.
[10] Friedman J,Tibshirani R,Hastie T.Additive logistic regression:a statistical view of boosting (With discussion and a rejoinder by the authors)[J].The Annals of Statistics,2000,28(2):337-407.
[11] Chen T Q,Guestrin C.XGBoost:a scalable tree boosting system[C]//Proceedings of the 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining.San Francisco,California,USA.ACM,2016:785-794.
[12] 徐猛，雷洪，何江一，等.利用改进XGBoost预测RH精炼终点钢水温度[J].材料与冶金学报，2023,22(5):437-443.(Xu Meng,Lei Hong,He Jiangyi,et al.Predicting the endpoint steel temperature ofRH refining using improved XGBoost[J].Journal of Materials and Metallurgy,2023,22(5):437-443.)
[13] 查伟，董艳伍，姜周华，等.基于GBDT算法真空自耗铸锭终点锰含量预测模型[J].中国冶金，2023,33(7):107-114.(Zha Wei,Dong Yanwu,Jiang Zhouhua,et al.Prediction model of end-point manganese content in vacuum consumable ingot based on GBDT algorithm[J].China Metallurgy,2023,33(7):107-114.)
[14] 刘福龙，郝良元，靳亚涛，等.基于CatBoost算法的高炉铁水硫含量预测[J].河北冶金，2023(6):31-36.(Liu Fulong,Hao Liangyuan,Jin Yatao,et al.Prediction of sulfur content of molten iron in blast furnace based on catboost algorithm[J].Hebei Metallurgy,2023(6):31-36.)
[15] Dhananjay B,Sivaraman J.Analysis and classification of heart rate using CatBoost feature ranking model[J].Biomedical Signal Processing and Control,2021,68:102610.
[16] 蔡明，孙杰，李培德，等.三种机器学习算法在回归应用中的对比分析[J].智能计算机与应用，2022,12(8):165-170.(Cai Ming,Sun Jie,Li Peide,et al.Comparative analysis of three machine learning algorithms in regression application[J].Intelligent Computer and Applications,2022,12(8):165-170.)
[17] 廖红文，周德龙.AdaBoost及其改进算法综述[J].计算机系统应用，2012,21(5):240-244.(Liao Hongwen,Zhou Delong.Review of AdaBoost and its improvement[J].Computer Systems & Applications,2012,21(5):240-244.)
[18] 李玉琦.复杂预测模型超参数优化研究[D].北京：北京邮电大学，2022.(Li Yuqi.Research on hyperparametric optimization in complex prediction model[D].Beijing:Beijing University of Posts and Telecommunications,2022.)
[19] Bergstra J,Bengio Y.Random search for hyper-parameter optimization[J].Journal of Machine Learning Research,2012,13:281-305.
[20] Hutter F,Kotthoff L,Vanschoren J.Automated Machine Learning:Methods,Systems,Challenges[M].Cham:Springer International Publishing,2019.
[21] Parker-Holder J,Rajan R,Song X Y,et al.Automated reinforcement learning (AutoRL):a survey and open problems[J].Journal of Artificial Intelligence Research,2022,74:517-568.
基本信息:
DOI：10.14186/j.cnki.1671-6620.2025.06.003
中图分类号:TP18;TF713
引用信息:
[1]李星彤,龚伟,李帝阅.基于Boosting算法的转炉终点预测模型[J].材料与冶金学报,2025,24(06):589-596.DOI:10.14186/j.cnki.1671-6620.2025.06.003.
基金信息:
国家自然科学基金项目(52274324)