The Predicted System of Silicon Content in Molten Iron Based on Modular Neural Networks

This paper discusses on a modular neural network to improve system control performance indicators,based on module neural network forecasting method of hot metal silicon content.The modular neural network predictive control strategy is applied to establish a modular neural network predictive model,according to the nature of the input physical quantities constituting the four neural network modules,and then the output from the neural network prediction of silicon content in hot metal predictive value,so as to control furnace temperature.Therefore,the simulation results show that the approach by using modular neural network prediction model has great advantages than general neural network prediction models for complex non-linear objects in improving the learning efficiency and generalized ability to respond effectively to improving the accuracy of the model estimates.Conclusively,the same amount has been divided by the modular neural network prediction hot metal silicon content in model,through the hot metal silicon content forecasting,which can improve furnace temperature control accuracy and dynamic tracking ability,with simple structure,good real-time forecast and high accuracy.