Neural Network Control And Analog Experimental Study On Hydraulic AGC System Of Cold Rolling Mill

For cold rolled strips, thickness accuracy is the leading indicator to judge the qualityof strip products, and the thickness control technology plays a vital role in the field of coldrolling. So deeply researching the thickness of the rolling mill control is very necessary, atthe same time, to apply some theory in modern control theory to the thickness of therolling mill control system also has important theory meaning and actual reference value.Based on the deep understanding of control theory of cold rolling mill hydraulic AGCsystem in the paper, the PID controller widely used at present is reformed. Traditional PIDcontroller for the strong coupling, nonlinear, time varying thickness control system hasmany limitations. therefore, this article chooses BP neural network which has the onlinefunction in the intelligent control to design a adaptive PID controller, and the artificial fishalgorithm is used to optimize BP network parameters off-line, combining the two methods,the thorough simulation and the experimental study are made.First of all, based on the mill bounce equation, both the bounce equation curve ofrolling mill and the rolled piece plasticity curve are combined to analyze the influence onthe export thickness of the rolled piece resulting from various interference factors.Combining rolling mill hydraulic AGC system with its operation mechanism, and on thebasis of the basic equation of servo valve, hydraulic cylinder continuity equation and forcebalance equation of both the hydraulic cylinder and load, a mathematical model forhydraulic AGC system of cold rolling has been established and an analog circuit of coldrolling hydraulic AGC system is set up according to the mathematical model.Secondly, a controller is designed based on BP network, and I have some experimentswith it. Analyzing the shortage which is easy to lost in local extremum in the BP network,the artificial fish algorithm is introduced to optimize the BP neural network.Finally, the software of FB Generator and the C language are used to write thepackage program for the optimized network, then an intelligent controller of BP neuralnetwork optimized by the improved artificial fish algorithm has been obtained, and CFC(Continuous Functional Charm) block is also created according to neural network intelligent algorithm. On the platform of Siemens FM458, the control experiments aremade for the model of cold rolling mill hydraulic AGC system.