Research On BOF Steelmaking Endpoint Control Model Based On Neural Network

The process of steelmaking is so complex, the environment is so harsh and the temperature is so high that it is very difficult to measure the temperature and contents continuously;therefore, we can't apply normal methods of process control on the blowing process. So it becomes very important to accurately predict the endpoint temperature and element contents. The prediction is the basis of endpoint control. Neural network (NN) is effective in solving nonlinear problems. Therefore, it is promising to apply neural network to the endpoint control of BOF (Basic Oxygen Furnace process). This paper briefly introduces some artificial neural network models that are used to predict the endpoint of BOF steelmaking, proposes a prediction model based on the RBF NN and a control model based on the BP NN of the endpoint steel temperature (or the endpoint carbon content). The main contributions of the thesis are as follows:A prediction model of endpoint steel temperature (or the endpoint carbon content) is designed by the advantage of the RBF NN that has quick convergence velocity and can effectively avoid the problem of local minimum. The selection of centers of RBF NN becomes the bottleneck that retards the broader application of RBF NN. This paper adopts nearest neighbor algorithm to select the centers of RBF NN, the centers of hidden layer are variable with the precision, the algorithm avoids the disadvantage of the K-means which relies on the initial central position and is possible to enter local minimum point. The simulation results show that the precision of prediction method of endpoint steel temperature (or the endpoint carbon content) is high.On the basis of Prediction model, make use of the ability of nonlinear reflection of the BP NN. An internal control model of endpoint steel temperature (or the endpoint carbon content) is design. The model can account the need of coolant and oxygen during the reblowing by the dynamic endpoint control method. The simulation result shows that the new method overcome the shortcoming that the control model based on the heat balance and the oxygen balance was not accurate in the traditional method, the endpoint hittingratio was raised.