BP neural network prediction for Si and S contents in hot metal of COREX process based on mathematical analysis and Deng’s correlation

Abstract

In COREX operation, the Si and S contents in hot metal are relatively high and easy-fluctuating, which is one of the problems affecting the practical operation. Accurate predictions of Si and S contents can provide theoretical references for stabilizing the fluctuations and decreasing the contents of Si and S in hot metal. Therefore, the present work established the prediction model of Si and S contents in hot metal in COREX based on BP neural network. The results show that the root-mean-square errors between the predicted value and actual value for Si and S are 0.098 and 0.0037, respectively. They are 0.070 and 0.0040 when the time-sequence lapse method is adopted, which turns out to be better. Therefore, the model is accurate and reliable to predict the Si and S contents in hot metal in COREX.