Research On Breakout Prediction System For Continuous Casting Beased On RBF Neural Network And Fuzzy Theory

Breakout is the most catastrophic accident in the continuous casting process. It is a combination of factors that deviate from the steady-state results of the various technology effects to outflow of molten steel,not only influence the personnel casualty and safe production, but also the quality of casting slab which will cause a great economic loss. Consequently, researching the formational mechanism of breakout and then exploring the new technologies to predict the breakout can reduce this catastrophic accident in the continuous casting process and be valuable in theoretical research as well as the societal economy.In the paper,the process of continuous casting ,the variety of breakout and the cause of the breakout in sticking casting are firstly analyzed and the preventive countermeasures are put forward. On the basis of this deeply research, a new multilevel network breakout prediction system is established, composed of RBF neural network and T-S fuzzy neural network to conduct breakout prediction.After that, RBF neural network is adopted to identify the temperature wave of single thermocouple. Improving the dynamic of the nearest neighbor clustering algorithm,the RBF network makes a better approximation ability and can get better network performance. According to the laying of the actual location of the thermocouple, using different combinations of space achieves a regional test based on transverse network, in order to identify the shell blank space in the breakdown lane on the expansion. In view of the thermocouple temperature wave vulnerable to the impact of casting process parameters, a mould is put forward, in which fuzzy neural network is utilized to consider the factors what arose the fluctuation of the steel liquid surface. The simulation results show that the breakout prediction system further reduce the false alarm rate and improve forecast accuracy.Finally, in WINDOWS platform, using MATLAB language to achieve the breakout prediction system's off-line simulation, the dangerous situation is given warning timely. At the same time, the system's visual interface is developed. It reflects the change of the thermocouple temperature and displays dangerous thermocouple's position. it achieves a good human-computer interaction.