Breakout Prediction System Based on Combined Neural Network in Continuous Casting

Abstract

A breakout prediction system based on combined neural network in continuous casting is developed. It adopts the radial basis function (RBF) neural network for single-thermocouple temperature pattern pre-diagnosis, and logic judgment unit for multi-thermocouple temperature pattern recognition at first, then uses fuzzy neural network based on Takagi-Sugeno (T-S) model to make final decision. In the RBF network, the maximum entropyfunction is used to normalize input data. According to the law of crack growth in the bonding location, the horizontal network prediction model is adopted to logically judge multi-thermocouple temperature pattern, the prediction time is shortened. In the T-S fuzzy neural network model, the overall influencing factors of breakout are considered. The results show that the breakout prediction system based on combined neural network can effectively decrease the false alarm rate and improve the prediction accuracy.