The Research On Cold-Strip Steel Flatness Intelligent Prediction Control

In this paper, the author aims to realize the flatness on-line feedback control. The author chooses the cold-strip steel flatness intelligent prediction control as research subject. According to the basal principle of prediction control, the cold rolling flatness intelligent prediction control system is established for 1220 five-stand four-high cold tandem mill.First, basic conception of flatness, representation of flatness, the common defective flatness and corresponding control law are explained. Then, many flatness pattern recognition methods are introduced. In this paper, according to the requirements of flatness on-line control, Least Square Method based on Legendre orthodoxy polynomial is chosen for flatness pattern recognition. The pattern recognition program is implemented by Matlab software, and recognition examples are given.Second, according to the existed flatness prediction models, the modified Elman network flatness prediction model is established. The optimization of network structure is done by genetic algorithm, and feedback compensation is used to the output of prediction model. So the prediction precision is improved. By Matlab software, the author realizes flatness prediction program. The simulation results show that, compare to the mathematic model and BP network model, this model is suit for actual rolling process, and has high precision.Third, the flatness on-line rolling optimization control model is set up with dynamic fuzzy neural network. The on-line training algorithm and off-line training algorithm based on GA-BP for this model are deduced. The flatness on-line rolling optimization control model program is written by Matlab software. The figures of membership functions are given respectively before and after optimization.In the end, the flatness intelligent prediction control program is compiled using Matlab software. The simulations of flatness on-line control are given combining examples. The results show that, the flatness intelligent prediction control system established in this paper can slake simple wave and quadratic wave. This study has important meaning on realizing flatness on-line real-time control and promoting the development of flatness control model.