Study On Multilevel Networks Breakout Prediction System For Continuous Casting

Breakout is reduced greatly, economy loss is depressed, and personnel casualty is reduced, due to the development of breakout prediction system, but leaking alarm rate and false alarm rate can not be eliminated. In the paper, two aspects of how to prevent the sticking happened and how to improve the accurate of the system are studied.The mechanism of sticking breakout is analyzed in the paper, educing that the essential reason of sticking is the damage of the liquid steel meniscus. Many factors influencing liquid steel meniscus are set forth, amend measures are put forward to prevent the sticking.On the basis of deeply research to many breakout prediction systems, a new multilevel network breakout prediction system is established, composed of BP neural network and T-S fuzzy neural network. BP neural network is adopted to identify the temperature wave pattern of single thermocouple. Curve fitting technical based on maximum-entropy function is put forward in disposal of input data of the network, and the network framework is reduced. Based on the extend rule of sticking solidified shell breaching, combining the actual instance, measure temperature in transverse is superior to what in longitudinal, transverse network is replaced by longitudinal network, the prediction time is shortened.The temperature pattern of thermocouples is often influenced by the fluctuation of the steel liquid surface in mold, and false alarm is formed. For the sake of reducing false alarm rate, 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 result showed that the false alarm rate of the system is greatly reduced.The whole system is developed by MATLAB language on the Windows platform, the visible interface is also developed, by which the dynamic display of the temperature wave of the thermocouple is realized, and the prediction result is showed in time, prompting the operator to take measures to avoid the breakout incident.