The Cold Rolling Strip Flatness Controlling Research Based On Neural Networks

Neural network control is a typical form of AI control. it has been developing at full speed in resent years. Especially to the complicated uncertain indefinite nonlinear systems with multi-inputs or multi-outputs that are hard to establish ideal mathematical model, neural network control theory is applied. And neural network control being realized in that complicated systems through computers usually can get a ideal controlling effect and the equipments being used are simple, remarkable results in economic benefits can yield.High quality Cold rolling product with good flatness is the goal of the domestic or foreign metallurgical industry. Using traditional method to control steel strip flatness ,we cannot obtain ideal effect , because mechanical organization change can not reflect to the control system promptly and make some revisions.This paper introduces flatness defect of cold rolling product, and in view of these different defects , analyzes how to eliminate these defects through adjusting mill actuators to obtain good strip profile. In the same time it states Neural network control's primary principle,construction,designing rules and the algorithms of self-adapting facts.On the base of the theory of Neural network control, a self-adapting neural network control system combining computer control technology and Neural network control technology is applied in flatness control process. Through the HMI man-machine interface, with the off-line way training neural network, enable it to have some recognizable ability, then carries on the on-line training in this foundation, gradually enables the control to achieve the perfect condition. Demonstrated through the HMI machine human interface, the nerve network control output can fast track the initial value, the real-time control effect and the stability and robustness are good. It can decrease the cost of production and increase the total profit. It is the future development direction of flatness control technology.