统计学方法分析CSP精轧压力模型  

Rolling load model of CSP finishing mill group based on statistics method

作　　者：叶奔[1,2] 李立新[2] 胡盛德[2] 

机构地区：[1]中冶南方工程技术有限公司热轧分公司,湖北武汉430223 [2]武汉科技大学耐火材料与冶金国家重点实验室,湖北武汉430081

出　　处：《钢铁》2015年第10期54-58,共5页Iron and Steel

摘　　要：采用统计学方法分析CSP的轧制数据,发现存在严重的自相关现象。结合工艺分析了自相关的成因,采用主成分分析法和岭回归法进行了分析以消除自相关性。通过对志田茂模型的拟合,验证了自相关的不利影响,并应用统计学对数据进行了部分剔除。拟合了志田茂改进模型和多元回归模型,精度得到提高。引入再结晶分数模型后,轧制力模型得到了较好精度。采用新的30个数据对5个模型进行了对比验证,结果显示考虑再结晶的模型5精度较好。Data of CSP finishing mill group were analyzed by statistical method. The analysis of autocorrelation was carried out and the presence of serious autocorrelation phenomenon was revealed. The principal components analysis and ridge regression method were occupied to eliminate the autocorrelation. By fitting the Shida Shige model,the effects of autocorrelation were verified. Serious statistics method is used to eliminate some data. Modified Shida Shige model and multiple regression model were built. Then the precision of models is improved. The introduction of recrystallization gets a better accuracy. In the comparative test for 5 models,a set of 30 new data was used. Results show that the recrystallization model 5 has better accuracy.

关 键 词：统计学 CSP 轧制润滑 预报 自相关 主成分法 

分 类 号：TG335[金属学及工艺—金属压力加工]