A regularization-based neural network for the measurement of shape information

The measurement and disposal of shape information plays a vital role in strip rolling process and it also is one of the focus in strip rolling theory today. In this paper, based on the predictive shape control system, a predictive model is developed in which the regularization method is adopted. Through the simulation on the data of four high reversible rolling mills, it proved this model can conquer the defects of the mathematical model effectively and get a desirable performance of shape control easily.