Research On Rbf Flatness Forecasting Model Based On Modified PSO Algorithm

Steel is the important material basis for developing the national economy and enhancing the comprehensive national strength, and the plate and strip is applied extensively as the important raw and processed materials in every department of national economy. Flatness is one of the quality indexes of plate and strip, the accurate flatness forecasting model is needed in no matter what the control characteristic analysis of adjusting framework in flatness control system, also the online real time control.So it is more and more urgently to build the accurate forecasting model.Analyzing on flatness forecasting model in the current situation, and finding the problems of the previous methods, a research on flatness forecasting model has been made in this paper.Firstly, researching on the PSO, in order to improve the limitation of easily getting into the local optimum in the basic PSO, a modified PSO is presented in this paper. The chaotic optimization algorithm is introduced to decide the parameters of PSO dynamically in the global space, and thereby it can solute the question of parameter dependent in the basic PSO. The experiment is done by using the sutra testing functions to test the performance of modified PSO.Secondly, the MPSO-RBF hybrid optimization strategy is presented. In order to improve the convergence rate and precision in the learning algorithm of neural network, the MPSO algorithm, which has global searching ability and strong practicability, is combined with the RBF neural network training. At the same time, the particle coding of MPSO algorithm, operation design and steps of hybrid optimization algorithm are presented. Take the simulation example by Hermit polynomial approximation and the Iris classification problem, which can test RBF neural network trained by PSO and MPSO, and compare their training precision and convergence rate.Finally, the flatness forecasting model based on MPSO-RBF neural network is built. After normalizing the practical statistical producing data, the influence factors of flatness during rolling process are used as the input, and the character coefficients of flatness are used as the output, then the forecasting model based on MPSO-RBF is put forward. And the simulation experiments of this model are realized by Matlab software.