Research On Decoupling Control Of Looper Based On Neural Network Inverse System

With the development of rolling technology,the requirements of modern hot strip mill in the field of strip products are more and more strict.The looper tension and height control largely determine the performance of steel products in hot rolling process.By using decoupling control method,the mutual interference between the two can be eliminated.So the control precision of the tension can be improved further,and the quality of the strip products can be improved.Therefore,Loopers' height and tension of decoupling control method is always in constant progress.Based on the background of 650mm hot rolling mill,the looper decoupling control research and analysis of the key,the main research achievements are as follows:First of all,the process of hot rolling and hot strip rolling are introduced in detail.Then,I analyze the causes of tension and establish a coupled mathematical model of looper height and tension.By using diagonal matrix method,the loopers' tension and height are decoupled and the simulation results are analyzed.Then,I research on the neural network inverse system theory and neural network inverse system method.I demonstrate the reversibility of the process and make the design procedure of neural network inverse system clearly.I also design the structure of neural network inverse system.Then I make the research on the neural network inverse system implementation steps and composite control method in order to lay the foundation for the following simulation.Finally,I analysis the reversibility of the system.Through the steps of selection of neural network,excitation signal selection,training the input and output sample of neural network off-line and generating neural network module to complete the loopers' neural network inverse system identification.The neural network inverse decoupling control system and its simulation platform are established.I decouple looper system in order to analyze the decoupling effect.At last,The method has a excellent result by MATLAB simulation.