基于BP神经网络的RH精炼终点钢液温度预测  

Predicted temperature of molten steel at the end of RH refining on the base of BP neural network

作　　者：柴宝堂 雷洪 徐猛 赵岩 CHAI Baotang;LEI Hong;XU Meng;ZHAO Yan(Key Laboratory of Electromagnetic Processing of Materials,Ministry of Education,Northeastern University,Shenyang 110819,China;School of Metallurgy,Northeastern University,Shenyang 110819,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,辽宁沈阳110819 [2]东北大学冶金学院,辽宁沈阳110819

出　　处：《炼钢》2023年第5期33-40,47,共9页Steelmaking

基　　金：国家自然科学基金与宝钢联合资助项目(U1460108)。

摘　　要：合理地控制RH精炼终点钢液温度,可以有效地保障连铸过程中拉速的稳定,明显提升铸坯的质量。为了精准预测RH精炼终点钢液温度,先采用Spearman相关系数分析各影响因素与RH精炼终点钢液温度之间的关系,筛选出相关系数较高的影响因素;然后运用SPSS软件进行主成分分析,得到一组互不相关且能较好地描述RH精炼终点钢液温度的影响因素;最后分别采用粒子最优化算法、遗传算法和退火算法优化BP神经网络模型预测终点温度。计算结果表明,到站温度、进站钢水氧值、真空时间、钢中初始碳含量、钢包净空、脱碳终点氧值这些因素是RH真空精炼终点钢液温度预报的主要因素。粒子最优化算法优化BP神经网络预测RH处理终点钢液温度的温度极差最大,为20℃。该模型的均方根误差和平均绝对误差最小,分别为3.38和2.35,且预测RH处理终点钢液温度±5℃的命中率最高,达到90.0%。Appropriately control of the temperature of molten steel at the end of RH refining can ensure the stability of casting speed during the continuous casting effectively,and improve the quality of billet significantly.In order to accurately predict the temperature of molten steel at the end point of RH refining,Spearman product-moment correlation coefficient was applied to analyze the relationship between the factors and the temperature of molten steel at the end of RH refining,and the factors with higher correlation coefficient were given.Then SPSS software was applied to do the principal component analysis in order to determine a set of independent factors to describe the temperature of molten steel at the end of RH refining.Finally,the particle optimization algorithm,the genetic algorithm and the annealing algorithm were applied to optimize BP neural network model to predict the temperature of molten steel at the end point of RH refining.Numerical results showed that,arrival RH temperature,initial concentration of oxygen in the molten steel in RH,vacuum time,initial concentration of carbon in the molten steel,head room of ladle,terminal concentration of oxygen in the molten steel in RH were the main factors to predict the molten steel temperature at the end of RH treatment.If BP neural network optimized by particle optimization algorithm is applied to predict the molten steel temperature at the end of RH treatment,the temperature range is the maximum,and up to 20℃.The value of RMSE and MAE of this model are the least,and up to 3.38 and 2.35 respectively.The hit rate of the molten steel temperature±5℃is the best,and up to 90.0%.

关 键 词：RH精炼 粒子最优化算法 遗传算法 退火算法 BP神经网络模型 钢液温度 

分 类 号：TF769.4[冶金工程—钢铁冶金]