Research And Simulation Of Hydraulic AGC System For Four Revesed Roller Finishing Mill In Tianjin Iron&Steel CO.

This dissertation establishes the dynamic model on the basis of analyzing the thickness control of the plate rolling mill with an actual project of the company. This dynamic model reflects the actual situation of the project with the coulomb friction and the centrifuge of the support roller taking into account. Meanwhile, we make the PID controller on the BP neural network according to the dynamic model.The application of BP neural network in PID control can effectively overcome ill parameter adjusting and poor performance when the plant is nonlinear, time-varying, uncertain, and difficult to set up an accurate mathematics model. This paper proposes a new type of PID control method based on BP neural network on line by applying the improved conjugate gradient algorithm.The simulation results show that this improved algorithm increases the convergence speed in training process, improved systematic robustness and dynamic performance with the help of better self-adaptation and self-learning of the improved trained BP neural network. In conclusion, the new design has shown effective system control and met the anticipated goal. And compared the actual system with the simulation wave, we can validate its truth. The result indicates the good performance of the PID controller.