The Research Of Neural Network Predictive Model For Hot Rolling Strip Width Based On Improved Particle Swarm Optimization

With the continuous development of science and technology, the proportion of complex industrial production processes in modern industry is growing. However, the complex industrial processes are often nonlinear, multi-loops, multi-layered; the complexity has brought great inconvenience to the actual production. Especially the construction of model in complex industrial processes is basis of production, scheduling and control, it is significant for improvements of industrial development level and safe stable operation. Therefore, to find a good suitable modeling method for complex industrial processes is particularly important. In recent years, the development of intelligent technology provides a new research direction for complex industrial process modeling.Supported by Natural Science Foundation of ShanXi Province (No.:2010011022-3), this paper researched and designed a neural network predictive model based on a modified particle swarm optimization. And taking the width of a hot rolling mill as study object, a forecasting model of a hot rolling mill width is established.The main works of this paper are:(1) The background and significance of this subject were elaborated.Based on the research of various modeling methods in the complex industrial processes, the research content and method of this study were determined.(2) The principles of particle swarm optimization, optimization process and algorithms execution flow were analyzed in detail, and the particle swarm optimization algorithm parameter selection and convergence were analyzed and summarized. According to the shortcomings of particle swarm optimization, an improved particle swarm optimization was proposed.(3) After the analysis of BP neural network topology and the learning process, aiming to the lack of BP neural network, the improved particle swarm optimization algorithm was adopted as the network parameters optimization algorithm, so particle swarm optimization and neural network were combined. A neural network modeling method based on improved particle swarm optimization was proposed and the optimization process was given.(4) The production process of hot rolling strip was studied and mastered after extensive research in a hot rolling steel production, and the factors impacting the striped steel width changes in production process was researched. The forecasting model for hot rolling strip width was established by the proposed method.(5) In the MATLAB simulation platform, a large number of simulations and studies are carried out about the improved particle swarm optimization algorithm and neural network based on improved particle swarm respectively, good results was achieved. In addition, the proposed neural network was trained by hot-rolling steel production actual datasets for the width prediction, and then the network was tested by test samples, the simulation results show that the method is feasible and effective.