Research On Decoupling Control Method Of Thickness And Tension In Cold Rolling Mill

In the cold rolling process,the thickness accuracy of steel is one of the most important indexes to evaluate the quality of strip steel products.High precision tension control is a necessary condition for high quality steel production,and both of them are of great significance.However,the dynamic coupling between tension and thickness during the cold rolling process hinders the independence of the individual variables,making the control of the rolling system more difficult and limiting the quality of the steel product.The main contributions and contributions of this thesis are as follows:Firstly,the feedforward AGC,the mass flow AGC method and the thickness measuring AGC in the automatic control system of cold tandem rolling are briefly described.The direct tension control,indirect tension control and compounding tension control is briefly described.With the mathematical model in corresponding reference,the coupling model of thickness and tension is established.This coupling model is a continuous model,which is used as the controlled object in the simulation to verify the effect of controller.Then,a learning algorithm of recurrent neural network is designed by using equivalent static network and EKF algorithm.Corresponding calculation formulas are deduced and the validity of the learning algorithm is proved.Next,based on the traditional model predictive decoupling control,the learning characteristics of recurrent neural network are used to extract the dynamic characteristics of the controlled object,and a model predictive decoupling control based on RNN is established.Finally,the dynamic coupling characteristics of thickness and tension coupling model are extracted by learning ability of recurrent neural network,and the decoupling control of model predictive control based on recurrent neural network is used to decouple thickness and tension coupling model,and the schematic diagram of this controller is given.The simulation results show that the recurrent neural network based model predictive decoupuling control can compensate the coupling between thickness and tension effectively.