Predictive Model of BOF Based on LM-BP Neural Network Combining with Learning Rate

The endpoint temperature and carbon content of molten steel cannot be measured timely or accurately due to the extremely high temperature in BOF, so it is very important to establish an accurate predictive model for them. Steelmaking process is a very complex nonlinear process, and therefore it is very difficult to build up an accurate math model for it. The precision of traditional models based on oxygen balance and thermal equilibrium theory or based on reproducibility theory is low, and hit rate for prediction is low too. In this paper, the method that combines neural network technique with traditional modeling technology is adopted to build up static and dynamic models for steelmaking process. On this basis, presetting model is modified by using neural network technique to implement optimal setting control for steelmaking endpoint.