Research Of Bof End-point Optimization Control Model

The task of this dissertation is a sub-topic task of converter steelmaking expert system, which is a key scientific research attack project of Zhejiang province. The main puopose of this dissertation is to research some questions existed in the Jiangsu yonglian steel and iron company such as low steelmaking automation, full personal experience in steelmaking control, low blowing end point hitting ratio. Based on analyzing blowing craft, blowing process and a number of field data, build a small scale converter optimization model together ANN.The dissertation introduces the domestic and foreign present status and development tendency of converter steelmaking endpoint control technology, especially the successful application of nerual network technology in the BOF endpoint control based on reading many domestic and foreign references. Simultaneously according to investigating and analyzing the existed condition and question in lianfeng steel and iron company, we establish a endpoint object model in the case of the first turing down is unsuccessful in order to research temperature and carbon of the endpoint. The object model has two parts: Prediction Model and Control Model. The dissertation emphatically elaborates the modeling principle and process of optimization of the prediction model and control model.The dissertation analyse the source of the modeling data, and then these data were pre-treated. Three methods are used to establish the prediction model of temperature and carbon of the BOF endpoint in the dissertation: Linear Regression, RBF NN, and CIPSP-BP. The first two methods are the traditional method of modeling, but in the process of optimization of RBF NN the dissertation adopts nearest neighbor algorithm to select the centers of RBF NN, the centers of hidden layer are variable with the precision, the algorithm avoid the possible to enter local minimum point. The third method is to improve BP algorithm using the CIPSO algorithm, which has the advantage of high-speed flight and population diversity. The actual data of continuous 60 batches from a converter of the Jiangsu yonglian steel and iron company are chosen as example, 9 input variables that influence temperature and carbon of the endpoint in the converter are determined.And we establish three-layers RBF prediction neural networks and three-layers BP prediction neural networks, and predict temperature and carbon of the endpoint in the converter. The simulation result shows that the prediction model based on the CIPSO-BP of the endpoint steel temperature or the endpoint carbon content has the quickest convergence speed and the highest precision in the three kinds of prediction models. Therefore, the dissertation finally choose it as the basis of the internal control model.On the basis of the prediction model, the dissertation then make use of the area optimization to improve the control model based on the heat balance and the oxygen balance, and then establish the endpoint control model based on the area optimization. The simulation result shows that the method overcome the shortcoming of the control model based on the heat balance and the oxygen balance was not accurate in then traditinal method, the endpoint hitting ratio was raised.The dissertation finally summarizes the work which did to this dissertation topic, points out deficiency, simultaneously forecasts the work which dissertation topic next stage has to do.