热带终轧温度计算的逐步回归分析模型

基于大量现场工程记录数据,考虑轧制速度、精轧入出口厚度、机架间水量等影响热带终轧温度的主要因素,采用逐步回归分析法进行热带终轧温度计算模型的多元线性回归建模,在模型系数未进行学习时采用线性模型进行模拟计算,将计算结果与工程记录数据进行比较,模型计算结果与记录数据之差小于15℃范围内的达70.1%,表明该模型计算效果能够满足终轧温度控制精度要求.针对线性回归模型的局限性,对该模型系数自学习方法进行了讨论.

Stepwise Regression Analysis Model to Calculate Finishing Temperature During Hot Strip Rolling

Based on the in-situ data logging in a steel works in China, a linear model was developed to calculate the finishing temperature during hot strip rolling by way of stepwise regression analysis. The main influencing factors on finishing temperature were taken into account in modeling, such as rolling speed, workpiece thickness at entry and exit of finish rolling and inter-stand cooling water flow rate. A simulative computation was therefore done according to the linear model before starting the self-learning of model regression coefficients, then the computational results were compared with actual data logging. It was found that 70.1% of the difference between them is less than 15°C, which implies that the computational results of the model can meet the precision control requirement for finishing temperature. Furthermore, because of the limitation of the linear regression model, the self-learning of regression coefficients of the model was discussed.