关键词: 中厚板, 厚度控制, 弹跳模型, 轧制力模型, 自学习算法

Abstract: According to the characteristics of plate rolling, a more reasonable mill spring model and rolling force model are developed for the reform project of 2 500 mm finishing mill in Nan Steel. Considering that the high precision of the mill spring model and its self-learning is independent of the rolling force model, the self-learning is implemented first in the mill spring model as modification. Then, the workpiece gauge at exit is computed by the modified mill spring model used to implement the self-learning in the rolling force model where the short-term self-learning is used to correct the workpiece hardness coefficients of different batches, whereas the long-term one is to correct those of different steel grades. In this way, the cumulative results of short-term self-learning become the data source for subsequent computation. The method are kept and listed in database for subsequent computation. The method proposed was applied to the Medium Plate Mill Nan Steel, by which the hit rate of plate gauge control was increased by 13.3%.