Study On RBF Artificial Neutral Networks Prediction Model In Plate Controlled Cooling Process

Under the background of controlled cooling system project in a certain medium and heavy plate plant, this thesis analyzes and researches the controlled cooling system. At one time, basic automation and process automation are used to improve quantity and quality. To improve the control precision of the controlled cooling system, a RBF neural networks model of heat transfer coefficient factor is built to optimize the control model of finish cooling temperature. To ensure the stability of control model, the RBF neural networks model of heat transfer coefficient factor is combined with heat transfer coefficient mathematical model. The results through MATLAB emulation illustrate that the finish cooling temperate is improved obviously.The main works and results are as follows:1) Laminar flow in the plant cooling system and major equipment described in detail based on the combination of the production process of commissioning on site, on the basis of laminar cooling automation systems in the cooling water flow control, micro-plate track, the order of opening and closing the cooling water, interlock control, data acquisition, processing and exchange of the core functions of the program design and implementation.2) The calculation process of the main models of the process automation system is described in detail, include preset model, preset model and self-learning model. The laminar cooling system of this company, can meet the need of worksite, and satisfactory control effect with high degree of accuracy has been attained by analyzing and researching data from worksite.3) To improve the control precision of the controlled cooling system, a RBF neural networks model of heat transfer coefficient factor is built to optimize the control model of finish cooling temperature. The working principle of the RBF network, the learning process was studied in detail. The deficiency of original RBF neural networks is improved, so that the efficiency and precision of RBF neural networks model are improved.4) The RBF neural networks model is trained and tested by MATLAB. The results show that RBF neural network self-learning model has a high training accuracy. Compared to coefficient exponential smoothing method, the self-learning forecasted by RBF network adjusts a lesser extent.5) According to the characteristic of RBF neural networks model and the worksite condition, in this paper, a feasible project that is developed. The RBF neural networks model of heat transfer coefficient factor is combined with heat transfer coefficient mathematical model to ensure the stability of control model, the emulation results illustrate that the finish cooling temperate is improved obviously.