热带钢精轧机组轧制力预设定模型自学习研究

Research on self-learning of hot strip finishing setup model for rolling force

针对某厂热连轧精轧机组预设定模型的自学习模块进行研究。在短期自学习方面,提出一种多变量控制的平滑系数模型,研究结果表明:所提出的模型优化效果明显优于以往单变量控制平滑系数模型;长期自学习方面,主要研究长期自学习的启动条件,以及换层别后首块钢学习系数的选取策略。在长期自学习启动条件中加入规格变化程度的判定条件,在保证预报精度的前提下,有效减少了长期自学习启动次数,保证了自学习的连续性。其次,在原自学习模型中加入趋势学习系数,在一定程度上修正了长期自学习系数中所包含的设备状态信息,提高了长期未轧制层别的轧制力预报精度。最后,优化了从未轧制过层别初始自学习系数的选取策略,通过对已轧层别中相似层别学习系数的学习,有效提高了从未轧制的层别的轧制力预报精度。

To solve the problem of self-learning of finishing model in some hot rolling, a deep research on the function of short-term self-learning and long-term self-learning models was made. The point in the short-term self-learning is how to premise the smooth modulus in the method of exponential smoothing. A multi-variable control exponential smooth model was proposed. The results show that the result obtained by the proposed model is more better than the result of only using the single modulus self-learning model. In the aspect of long-term self-learning, the main mission is the launch conditions of the self-learning, and how to select the strategy of learning coefficient about the first piece of steel after changing the layer. Firstly, the judge conditions about the degree of the changing of the standard is joined into the launch conditions of the long-term self-learning, the forecast accuracy is ensured, it effectively reduces the launch times in long-term self-learning, the continuity of the self-learning is improved. Secondly, the method of tendency learning modulus is put into the original rolling force learning model. It effectively renews the equipment message belonged to the self-learning modulus in layer table, improves the forecast accuracy in the layer never rolled. Finally, the strategy of the self-learning modulus in the layer never rolled is proposed by using the similar self-learning modulus in the rolled layer, which effectively improved the accuracy of the initial self-learning modulus in the never rolled layer.