Research On Flatness On-Line Intelligent Control For The Wide Strip Steel Cold Mill

Shape control is the key technique and important develop direction for modern times high precise wide strip steel cold mill. Based on the theory of artificial intelligence, author chooses flatness on-line intelligent control for the wide strip steel cold mill as research object with theoretical and engineering sense, makes embedded theoretical research and industrial experiment as well as application, and has achieved new fruit.Firstly, in order to overcome the defects of the traditional models for flatness recognition, the RBF (Radial Basis Function) network based on SVM (Support Vector Machine) model has been applied, so that an effective model of flatness pattern recognition based on RBF network has been constructed. This method based on the structural equivalence of the SVM and the forward network, the SMO (Sequential Minimal Optimization) algorithm is employed to obtain more optimal structure and initial parameters of neural network, and then the BP (Back Propagation) algorithm was used to adjust the neural network. This solves the problem in ascertain the basis function and the hidden layer's parameter of the RBF network and new method is proposed for building the model for flatness on-line recognition and forecast.Secondly, because of the deficiency of the traditional static effective matrix method and the analysis of regulative capacity to the flatness, the transfer effective matrix method for flatness control has been founded. This method indicated that the effects of the flatness control measure on the flatness are different with the difference of regulative ranges. The transfer effective matrix must be calculated on line. So the influence of every regulative measure can be reflected correctly, and the new regulation can be conformed.The flatness transfer effective matrix model based on clustering and fuzzy neural network has been founded. By introducing the fuzzy clustering into the fuzzy neural network modeling, the number of the nodes is confirmed by the number of the clustering center, so that the numbers of the rule are reduced. The optimal structure can make the flatness transfer effective matrix model meet the real-time need. The law created automatically can help guiding the rolling production process. The model has been applied on the 1220 cold rolling mill for industrial experiment to assess the benefits offered by the transfer effective matrix flatness control method based on the fuzzy neural network model, and the frame of flatness transfer effective matrix control system has been designed.Finally, because the rolling load forecast is the important influencing factor to flatness in temper process, the research on rolling force forecast has been made. The method of rolling load forecast combined neural network for two-stand temper mill has been produced. According to the least elongation for eliminating the flatness flaw, the model of elongation distribution modulus for two-stand temper mill has been designed. A friction coefficient model for the two-stand temper mill has been proposed. The massive academic foundation has been established for developing the rolling preset system. The neural network model of deformation resistance has been founded to solve the problem that the deformation resistance can be obtained only by the experiment. Combining the rolling theory, the theory model, self-adapting, self-learning of elongation distribution and the neural network, the model of rolling load forecast combined neural network for two-stand temper mill has been founded. The method has been applied on the 1220 two-stand temper mill for industrial experiment to access the rolling load model. Therefore, a new system framework has been designed in the original computer control system. The program of rolling load forecast has been developed and the experimental result has been analyzed.Choosing flatness on-line intelligent controlling model of wide strip steel cold mill with theory and engineer practice meaning as research object, author has done a lot of theoretical, simulating and industry experiment research on flatness pattern recognition, on-line forecast, on-line control and preset. This not only has the research important significant for developing flatness recognition and control theory, but also the practical sense and applying value for the improvement of on-line high precise flatness control and the development of technology of the flatness control.