数据驱动弹簧钢脱碳影响要素分析与质量预测  

Data driven analysis of influencing factors of decarburization and quality prediction of spring steel

作　　者：徐佳昀 吕立华[2] 蒋嘉石 施逸非 姜庆超[1] 颜学峰[1] XU Jiayun;Lü Lihua;JIANG Jiashi;SHI Yifei;JIANG Qingchao;YAN Xuefeng(Key Laboratory of Smart Manufacturing in Energy Chemical Process,Ministry of Education,East China University of Science and Technology,Shanghai 200237,China;Intelligent Institute of Central Research Institute,Baoshan Iron and Steel Co.,Ltd.,Shanghai 201900,China)

机构地区：[1]华东理工大学能源化工过程智能制造教育部重点实验室,上海200237 [2]宝山钢铁股份有限公司中央研究院智能所,上海201900

出　　处：《冶金自动化》2022年第5期103-111,共9页Metallurgical Industry Automation

基　　金：国家自然科学基金面上项目(61973119,21878081);上海市青年科技启明星项目(20QA1402600)。

摘　　要：脱碳质量对于线材钢产品质量和性能有显著影响,因此提出了一种基于线材生产过程数据的弹簧钢脱碳影响要素分析与质量预测方法。首先,采用XGBoost进行变量选择;其次,采用沙普利加法解释模型(SHapley Additive exPlanation,SHAP)对影响弹簧钢质量的要素进行分析与解释;再次,采用偏相关分析(partial cross mapping,PCM)构建影响要素的因果关系图,对影响脱碳质量的根因进行识别;最后,基于XGBoost构建脱碳质量状态预测模型。利用宝钢高速线材产线弹簧钢脱碳数据进行方法验证,选出了12个脱碳影响要素,在此变量选择结果上分别使用SHAP和PCM方法获取了更多的数据特征信息,利用该变量子集构建XGBoost脱碳质量预测模型,准确率为88.72%,检出率可达到93.89%,与使用全体变量建模的结果接近,验证了所提方法的有效性。Decarburization quality has a significant impact on the quality and properties of wire rod steel,so based on the data of wire rod production process,an analysis method of decarburization influencing factors and quality prediction of spring steel was proposed.Firstly,XGBoost was employed for feature selection.Secondly,the factors influencing the quality of spring steel were analyzed and explained by using SHAP.Thirdly,the causal diagram of influencing factors was constructed by using PCM,and the root cause of decarburization quality was identified.Finally,the prediction model of decarbonization quality state was constructed based on XGBoost.The method was verified by using the decarburization data of spring steel in Baosteel high speed wire rod production line,12 decarburization influencing factors were selected,and more data characteristic information was obtained by using SHAP and PCM methods,using this variable quantum set to construct the XGBoost decarbonization quality prediction model,the accuracy is 88.72%,and the detection rate can reach 93.89%,which is close to the result of using all variable modeling,verify the effectiveness of the method.

关 键 词：弹簧钢 脱碳 数据驱动建模 可解释性相关分析 偏相关分析 

分 类 号：TF762.5[冶金工程—钢铁冶金]