1.北京科技大学冶金与生态工程学院, 北京 100083
2.北方工业大学机械与材料工程学院, 北京 100144
3.江阴兴澄特种钢铁有限公司, 江阴 214429
4.中国第一重型机械股份公司, 齐齐哈尔 161042
樊士茜(2001—),女,硕士; E-mail:fanshiximetal@sina.com
段豪剑(1990—),男,博士,副教授; E-mail:duanhaojian@ustb.edu.cn
收稿:2025-03-25,
纸质出版:2026-01-30
移动端阅览
樊士茜,段豪剑,谢忠研等.基于SMOGN-XGBoost的钢包下渣剩余钢水量预测[J].特殊钢,2026,47(01):136-144.
Fan Shixi,Duan Haojian,Xie Zhongyan,et al.Prediction of Remaining Molten Steel Volume during the Process of Ladle Slag Carry-over Based on SMOGN-XGBoost[J].Special Steel,2026,47(01):136-144.
樊士茜,段豪剑,谢忠研等.基于SMOGN-XGBoost的钢包下渣剩余钢水量预测[J].特殊钢,2026,47(01):136-144. DOI: 10.20057/j.1003-8620.2025-00077.
Fan Shixi,Duan Haojian,Xie Zhongyan,et al.Prediction of Remaining Molten Steel Volume during the Process of Ladle Slag Carry-over Based on SMOGN-XGBoost[J].Special Steel,2026,47(01):136-144. DOI: 10.20057/j.1003-8620.2025-00077.
钢包结构直接影响炼钢工艺的效率、质量和经济性。为进一步优化钢包结构设计,基于钢包下渣水模拟数据,深入探讨了不同机器学习算法在预测开始下渣时剩余钢水量的效能,并针对钢包底部结构变量对下渣剩余钢水量的影响进行了预测分析。首先,采用SMOGN技术对钢包下渣水模拟数据进行过采样预处理,以平衡数据分布,构建包含训练集和测试集的剩余水量特征集。在此基础上,分别测试了LASSO,SVR,ElasticNet,MLP以及XGBoost五种机器学习模型对剩余水量的预测能力。通过决策系数、均方误差和平均绝对误差三个指标进行评估,结果表明,XGBoost模型的预测效果最优,是剩余钢水量预测模型的首选。最后,采用XGBoost模型分析了钢包模型底部结构变量,包括水口直径、水口凸起高度、钢包底部台阶高度和钢包底部台阶与水口距离等对钢包下渣剩余水量的影响。结果表明,当水口直径超过
<math id="M1"><mi>ϕ</mi></math>
https://html.publish.founderss.cn/rc-pub/api/common/picture?pictureId=98044989&type=
https://html.publish.founderss.cn/rc-pub/api/common/picture?pictureId=98045023&type=
2.03200006
2.87866688
40 mm 时,剩余水量显著降低。降低水口凸起高度,以及增加钢包底部台阶高度,会显著降低钢包内剩余水量:当水口凸起高度超过26 mm时,剩余水量则将超过20 L;而当台阶高度超过11 mm且水口凸起高度低于11 mm时,剩余水量将减少到10 L以下。当台阶与水口距离增大时,剩余水量先减少,在距离大于100 mm后趋于稳定。研究结果为钢铁企业优化钢包结构、降低钢液浪费方面提供了重要参考,具有实际指导意义。
The structure of the ladle directly affects the efficiency, quality, and economic aspects of the steelmaking process. To further optimize the ladle structure design, the current study deeply explored the effectiveness of different machine learning algorithms in predicting the remaining molten steel volume at the onset of the process of ladle slag carry-over based on water modeling data. Predictive analysis was conducted to explorer the impact of bottom structural variables of the ladle on the remaining molten steel volume during the process of ladle slag carry-over.Firstly, the SMOGN technique was employed to preprocess the water modeling data through oversampling, aiming to balance the data distribution and construct a feature set for residual steel volume, which included both training and testing datasets. Based on this foundation, the predictive capabilities of five machine learning models, namely LASSO, SVR, ElasticNet, MLP, and XGBoost, for the residual steel volume were tested. The evaluation was carried out through three metrics: the coefficient of determination, mean squared error, and mean absolute error. The results revealed that the XGBoost model outperformed the others in predictive accuracy, establishing it as the preferred model for forecasting the residual steel volume. Finally, the XGBoost model was utilized to analyze the impact of bottom structural variables of the ladle model on the residual steel volume. These variables included the diameter of nozzle, the height of nozzle, the height of steps, and the distance between nozzle and steps. The results indicated that when the nozzle diameter exceeded
<math id="M2"><mi>ϕ</mi></math>
https://html.publish.founderss.cn/rc-pub/api/common/picture?pictureId=98045000&type=
https://html.publish.founderss.cn/rc-pub/api/common/picture?pictureId=98044990&type=
2.03200006
2.87866688
40 mm, there was a significant reduction in the residual steel volume. Reducing the height of nozzle and increasing the height of steps significantly decreased the residual steel volume in the ladle: when the nozzle height exceeded 26 mm, the residual volume surpassed 20 liters; whereas, when the step height exceeded 11 mm and the nozzle height was below 11 mm, the residual volume was reduced to less than 10 liters. As the distance between nozzle and steps increased, the residual steel volume initially decreased and then stabilized after the distance exceeded 100 mm. The findings of this research provided significant reference for steel enterprises to optimize ladle structure and reduce molten steel wastage, offering practical guidance for the industry.
Zhang L F , Thomas B G , Wang X H , et al . Evaluation and control of steel cleanliness - Review [J]. 85 th Steelmaking Conference Proceedings , 2002 , 85431 - 452 .
Zhang L , Thomas B . State of the art in evaluation and control of steel cleanliness [J]. Isij International , 2003 , 43 ( 3 ): 271 - 291 .
Li Z W , Ouyang W , Wang Z L , et al . Physical simulation study on flow field characteristics of molten steel in 70 t ladle bottom argon blowing process [J]. Metals , 2023 , 13 ( 4 ): 639 .
Jardón-Pérez L E , Conejo A N , Amaro-Villeda A M , et al . Analysis of the Effect of Gas Injection System on the Heating Rate of a Gas Stirred Steel Ladle Assisted by Physical Modeling and PIV-PLIF Measurements [J]. ISIJ International , 2023 , 63 ( 3 ): 484 - 491 .
王家辉 , 张华 , 方庆 , 等 . 顶旋型湍流抑制器优化中间包流场的物理模拟 [J]. 钢铁 , 2023 , 58 ( 2 ): 72 - 82 .
赵烁 , 卢中阳 , 张泽 , 等 . 双水口钢包浇注末期汇流旋涡形成的影响因素 [J]. 钢铁 , 2024 , 59 ( 5 ): 71 - 79 .
杨晓江 , 周泉林 , 张全 , 等 . 复吹转炉渣金间传质物理模拟及应用 [J]. 钢铁 , 2022 , 57 ( 12 ): 57 - 65 .
Yin Y B , Yang J H , Zhang J M , et al . Physical modeling of slag carryover in the last stage of ladle teeming during continuous casting with dynamic change of slide gate opening [J]. Journal of Materials Research and Technology , 2023 , 23 : 1781 - 1791 .
Merder T , Kozłowski S , Pieprzyca J , et al . Physical modeling of two-phase liquid–gas processes occurring in the refining ladle for Fe–Si alloy refining process [J]. Scientific Reports , 2024 , 14 : 17565 .
李 嘉 , 刘 雨 , 王长军 , 等 . 气雾化制备18Ni250钢粉体的数值模拟与试验验证 [J]. 钢铁 , 2024 , 59 ( 6 ): 112 - 121 .
陈宏亮 , 刘珍童 , 周秋月 , 等 . 中间包内钢液流动和二次氧化的数值模拟 [J]. 中国冶金 , 2023 , 33 ( 7 ): 40 - 50 .
Li L , Tan D P , Yin Z C , et al . Investigation on the multiphase vortex and its fluid-solid vibration characters for sustainability production [J]. Renewable Energy , 2021 , 175 : 887 - 909 .
张 振 , 唐 珏 , 储满生 , 等 . 高炉冷却板挂渣能力数值模拟分析 [J]. 钢铁 , 2024 , 59 ( 11 ): 54 - 64 .
吴正义 , 武豪 , 江雪婷 , 等 . 浇注末期堵流操作对中间包内钢液特性的影响分析 [J]. 钢铁 , 2024 , 59 ( 8 ): 58 - 69 .
Zhang L F , Thomas B G . Numerical simulation on inclusion transport in continuous casting mold [J]. Journal of University of Science and Technology Beijing, Mineral, Metallurgy, Material , 2006 , 13 ( 4 ): 293 - 300 .
Duan H J , Huang C D , Zhang L F . Numerical simulation of transient flow and heat transfer in a steel ladle during holding period [J]. Metallurgical and Materials Transactions B , 2024 , 55 ( 4 ): 2273 - 2288 .
Huang C D , Duan H J , Zhang L F . Modeling of motion of inclusions in argon-stirred steel ladles [J]. Steel Research International , 2024 , 95 ( 11 ): 2300537 .
Ji J H , Li D Q , Du H X , et al . Optimize the operation parameters in surging auxiliary slag skimming process based on large eddy simulation–volume of fluid–discrete phase model coupled model [J]. Steel Research International , 2024 , 95 ( 3 ): 2300417 .
Thomas R N , Gupta R . A survey on machine learning approaches and its techniques: [C]. 2020 IEEE International Students' Conference on Electrical, Electronics and Computer Science . 2020 : 1 .
Li Y F , Chang J t , Kong C , et al . Recent progress of machine learning in flow modeling and active flow control [J]. Chinese Journal of Aeronautics , 2022 , 35 ( 4 ): 14 - 44 .
Zhang R H , Yang J . State of the art in applications of machine learning in steelmaking process modeling [J]. International Journal of Minerals, Metallurgy and Materials , 2023 , 30 ( 11 ): 2055 - 2075 .
李维刚 , 刘玮汲 , 谢璐 , 等 . 基于图卷积网络的热轧带钢轧制力预测 [J]. 钢铁 , 2023 , 58 ( 3 ): 89 - 96 .
梁印 , 朱航宇 , 罗林根 , 等 . 基于深度学习钢中非金属夹杂物图像识别 [J]. 钢铁 , 2023 , 58 ( 12 ): 62 - 70 .
张振 , 唐珏 , 储满生 , 等 . 基于EEMD和机器学习的烧结矿FeO成分长短期综合预报 [J]. 钢铁 , 2023 , 58 ( 8 ): 32 - 40 .
Zhang H T , Fu H D , Li W D , et al . Empowering the sustainable development of high-end alloys via interpretive machine learning [J]. Advanced Materials , 2024 , 36 ( 48 ): 2404478 .
Wentzien M , Koch M , Friedrich T , et al . Machine learning-based prediction of the martensite start temperature [J]. Steel Research International , 2024 , 95 ( 10 ): 2400210 .
Zhou K X , Lin W H , Sun J K , et al . Prediction model of end-point phosphorus content for BOF based on monotone-constrained BP neural network [J]. 钢铁研究学报:英文版 , 2022 , 29 ( 5 ): 751 - 760 .
Mitra R , McGough S F , Chakraborti T , et al . Learning from data with structured missingness [J]. Nature Machine Intelligence , 2023 , 5 ( 1 ): 13 - 23 .
Branco P , Torgo L , Ribeiro R . SMOGN: a pre-processing approach for imbalanced regression [C]. First international workshop on learning with imbalanced domains : Theory and applications[C]. Ljubljana: CEUR-WS , 2017 : 36 .
Tibshirani R . Regression shrinkage and selection via the lasso [J]. Journal of the Royal Statistical Society: Series B (Methodological) , 1996 , 58 ( 1 ): 267 - 288 .
Frank L , Friedman J . A statistical view of some chemometrics regression tools [J]. Technometrics , 1993 , 35 ( 2 ): 109 - 135 .
Zou H , Hastie T . Addendum: Regularization and variable selection via the elastic net [J]. Journal of the Royal Statistical Society: Series B (Statistical Methodology) , 2005 , 67 ( 5 ): 768 .
Thissen U , Pepers M , B Ü , et al . Comparing support vector machines to PLS for spectral regression applications [J]. Chemometrics and Intelligent Laboratory Systems , 2004 , 73 ( 2 ): 169 - 179 .
Hornik K , Stinchcombe M , White H . Multilayer feedforward networks are universal approximators [J]. Neural Networks , 1989 , 2 ( 5 ): 359 - 366 .
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 . 2016 : 785 - 794 .
Lundberg S , Lee S I . A unified approach to interpreting model predictions [EB/OL]. 2017: 1705 . 07874 . https://arxiv.org/abs/1705.07874v2 https://arxiv.org/abs/1705.07874v2 .
0
浏览量
3
下载量
0
CSCD
0
CNKI被引量
关联资源
相关文章
相关作者
相关机构
京公网安备11010802024621