Carbon And Temperature Endpoint Prediction Model Of Converter Steelmaking Based On RBF Neural Network

The control of carbon and temperature at the end point is an important link in the late stage of converter steelmaking.Because of super high temperature in the smelting process,it is difficult to accurately and timely measure,and hard to form a feedback control of common sense.By the limit of the structure size and investment&maintenance of equipment,most of the domestic small and medium-sized converter has not equipped with auxiliary lance and gas analysis apparatus,it is still in the situation that the detection equipment is far from perfect and automation control level is generally low.It is still a common phenomenon that the endpoint control process depend too much on artificial experience or traditional static model of steelmaking.According to the actual steelmaking situation on medium-sized converter from Ansteel Group,the method of pour converter and sampling is adopted mostly in endpoint control process.The endpoint hit rate is still very low.This is a great restriction on the quality of molten steel.Converter steel-making is a very complex physical and chemical reactions of the process,it is a subject with realistic significance that how to predict the endpoint carbon and temperature of converter steelmaking more precisely,real-timely,effectively and economically.The thesis's mainly research object are end point carbon content and temperature,the following main research work has been completed.(1)According to material equilibrium and hot equilibrium principle,under certain hypothesis,the converter smelting process is detailed analysised,the mechanism model about converter steelmaking endpoint prediction is built,as a result,the precision of the smelting model has been improved.(2)The RBF neural network is introduced to predict the endpoint control of converter steelmaking process.The RBF neural network is improved,it based on the nearest neighbor clustering algorithm and recursive least square method with forget--ting factor,and put forward the selecting method of sliding window based on the sample of matrix,establish the end point prediction model of converter steelmaking,which based on the existing data resources,the economic requirements and the condition of the medium converter.(3)Taking into account the smelting mechanism,the actual situation at the scene and the data requirements of model,the related smelting data were processed,there are mainly about outlier data processing and Standardized data processing.(4)Simulate the prediction model under MATLAB environment,collect steelmaking history data from Ansteel Group,train the simulation system,and applied this model to the prediction of end point carbon content and temperature of converter steelmaking.As the simulation results show,in the accuracy of ±0.019%,the C endpoint hit rate reaches 92%,in the accuracy of ±15?,the end temperature hit rate reaches 79%.The double hit rate for carbon and temperature reaches 71%.Compared with the current experience prediction method of Ansteel Group,carbon and temperature prediction results based on this model has been improved significantly,the model is proved to be effective,it can provide guidance for smelting operation conditions.