Flatness Pattern Recognition And Intelligent Control Via GA-PIDNN

With the development of economy and the progress of science and technology, plateand strip steel have been integrated into the development of social from all walks of life,and the requirement about the quality of strip is becoming higher. Flatness is one of theimportant indicators to test the quality of plate strip, flatness control is the key technologyof strip rolling mill. For many problems that exist in flatness control, the combination ofmodern control theory and intelligent control theory has become the development trend offlatness control. This paper chooses the flatness pattern recognition and intelligent controlas research subject, based on the intelligent control theory, PID neural network (PIDNN)model optimized by genetic algorithm (GA) was designed, using GA and BP algorithm tooptimize PIDNN respectively, in-depth comparative study was carried out and the flatnessrecognition and control have been realized.First of all, based on the analysis of the present flatness recognition methods andintelligent control methods, for the existing problem of low accuracy and it is difficult toobtain accurate model, PIDNN was introduced to the flatness recognition and control.Secondly, the structure and algorithm of PIDNN was studied, for the shortages of thetraditional BP algorithm in the process of optimization of PIDNN, GA was used tooptimize the PIDNN, so GA-PIDNN dynamic model was put forward. PIDNN not onlyhas the characteristics of traditional neural network, the proportional, integral anddifferential neurons with dynamic characteristics are also included in the hidden layer. GA,instead of BP algorithm, is used to optimize neural network, it takes full advantage ofglobal optimizing characteristics of GA and overcomes the shortage of easy trapped inlocal minimum of BP algorithm.In addition, for the900HC reversible cold rolling mill, flatness recognition modeland flatness predictive model are established via GA-PIDNN, and flatness intelligentcontrol system is designed. The research of simulation show that the accuracy of flatnessrecognition and predictive and the control effect based on PIDNN optimized by GA issuperior to BP algorithm. GA-PIDNN model has been fully utilized in flatness recognitionand control, and it is an effective flatness recognition and control method. Finally, for the quadratic component of flatness defect, a900HC four-high millhydraulic roll bending control system based on GA-PIDNN was proposed, the quicknessand effectiveness of GA-PIDNN controller is shown by simulation, and it is a goodmethod for the control of roll bending.