Research On Combination Of Intelligent Modeling For Hot-Rolling Strip Width

Iron and steel production is the main industry of the country. And hot-rolling strip production is the key development areas of the iron and steel production. With the development of market economy, customers request a higher demand for product specifications and quality. In order to meet market demand, customer requirements, and improve the competitiveness of the enterprise, the company must continue to optimize the control system, improve the product.The hot-rolling strip production process is a typical complex industrial process with the characteristics of strongly nonlinear, time-varying parameters and uncertainties. And with the complexity of the inter-linkages among production equipments, the existing mechanism model is difficult to meet the products requirements of the customers’high quality. A new modeling method of the combination of intelligent algorithms and conventional mechanistic model has become the new modeling direction of complex industrial processes.Supported by Natural Science Foundation of ShanXi Province (No.:2010011022-3), for the complexity of hot-rolling strip production process, using hot-rolled strip width control process as a research object, this paper design an intelligent model combining genetic-neural networks and width control mechanism model. And for solving the difficulties to establish the width model of small quantities and special steel, the paper designs case-based reasoning modeling.The main research contents for the paper are:(1) The description is given of the background and significance of the research topics. The research statuses of modeling of complex industrial processes are summarized at home and abroad. And the advantages and disadvantages of the width control model of hot-rolled strip are analyzed.(2) The production process of hot-rolling strip was studied after extensive research in a hot rolling steel production. And mastered width control mechanism model, control ideas, reasons that affect the accuracy of strip width and main factor of the hot-rolling strip.(3) An intelligent model combining genetic-neural networks and hot-rolled strip width control mechanism model was designed. Firstly, the BP neural network topology and learning algorithm were studied. For the defects of easily trapping into local minima, genetic algorithm was adopted to optimize the neural network weight and thresholds, and their work flow chart was given. Then using genetic neural network to optimize the width mechanism model parameters which were adjust by operators. Finally, a simulation study using actual production data and the result of statistical analysis were proved the effectiveness and feasibility of the proposed method. In the simulation process, the average impact value (MIV) was used to determine the genetic-neural network input. For the blindness of the initial parameter selection, the uniform design experiment was used to determine the initial parameters of the genetic-neural network, including population size, crossover probability, mutation probability, and the hidden layer nodes.(4) Using case-based reasoning approach to establish width model of small quantities and special steel. Firstly, the case base of small quantities and special steel in the previous production was constructed. Then the solution of new case was got by searching the case base, reusing, and modifying the solution of similar cases. Finally complete self-learning function by saving and maintaining case.