Research On Intellectual Control Strategy Of Flatness

This paper selects intelligent control strategy of flatness as research thesis based on artifical intelligent theory, makes embedded study on control strategy which combined pattern-recognition using neural network with fuzzy control of flatness, which has made certain achievement.In the research on neural-fuzzy control strategy of flatnees, this paper put forward a pattern-recognition method for flatness defect based on BP network optimized by Genetic Algorithm and design a flatness fuzzy controller on the basis of patter-recognition results. First, a practical GA code scheme is put forward. Moreever, genetic parameters are improved and topological structure of BP network is optimized. Based on these work, this paper constitutes six essence patterns of flatness defect using Lerande orthodoxy polynomial and design a GA-BP flatness defect pattern-recognition network model with 6 inputs and 3 outputs. Then, patter-recognition and controller are united into one function and it recognize the membership of flatness defect relative to six essence patterns as the forepiece of fuzzy controller directly which realized evaluation function of membership. Through the analysis of flatness defect characters, it define the fuzzy set rationally, deduce flatness fuzzy control algorithm and reduce calculational amout of fuzzy inference greatly.Through the simulation with field-measured data, it shows that GA-BP pattern-recognition method can recognize essence patterns from actual flatness defect accurately and has not only significant physical definitude, also high discriminating precision that reach 10-5 and satisfied effect. At the same time, it compared recognition effects with CM AC recognition model. The result shows that GA-BP has higer discriminating precison than CMAC, but CMAC achieve better in training speed, so these two method have their own strong point each. According to two performance index of flatness defect extracted in the paper, simulation result for flatness fuzzy controller indicates that this controller can control flatness defect to expected target fleetly and maxmiun defect can be restricted under ±2I, whichmeans that it has upstanding control performance.In view of that flatness and thickness control sysem is a coupled and complicated system, flatness control is effected and restricted by thickness control. So this paper also take research on flatness and thickness synthetical predictive control strategy. Through the analysis on mathematic model of flatness and thickness control system, it adopt GA-BP network to construct predictive model and on this basis it used centralized optimization-centralized control algorithm to deduce online controlling law for optimal rolling force and work-roll bending force. Also it adopt feedback modification method to revise control model online. Simulation shows that this strategy can realize decoupleing for flatness and thickness control and its control effects is better than feedforward compensation decoupling and PID control method, which establish the foundation of solving difficult problem that routine control method can't effectively control flatness and thickness synthetical system.