Control And Prediction Of Endpoint Oxygen Content Of Combined Blowing Oxygen Converter

According to the smelting condition of a certain steel mill 90t combined blowing converter, the article made the statistics and the correlate analysis on the base of the actual smelting data, found out the factor affecting to the molten steel endpoint oxygen content by the metallurgy thermodynamic analysis. An experience model of oxygen content was built during linear fitting, and a predictive model of terminal oxygen content of combined blowing oxygen converter based on MATLAB neural network was also developed.The results showed that the main influent factor of endpoint oxygen content was the molten steel carbon content, others were the manganese content, the phosphorus content and the molten steel temperature. The endpoint oxygen content increased along with the increase of the ferric oxide content of slag. The average carbon oxygen product in steel of smelting endpoint was 0.00259,it was bigger than theory calculation value 0.0024. The carbon content of 0.0339% and 0.0264% were the balance point for the [Fe]-[C] selective oxidizing when the temperature is 1,600℃ and 1,675℃. Critical carbon content of the [Fe]-[C] selective oxidizing drop when steel temperature increment.The latter stirring operation made the reaction between the steel and the slag approach the balance, reduced oxidability of the steel and the slag, removed the gas and the inclusion, and caused the ingredient and the temperature of the molten bath is evener. The influent factor of the oxygen content in the steel was argon current capacity, molten steel initial oxygen content as well as molten steel temperature in turn. The deoxidizing efficiency of 90t combined blowing converter was between 0.63 and 0.86.Because the process of steelmaking is a synthesis of multi-substance, multi-phase high temperature and attended by many physical chemistry reactions, and there are serious non-linear relations between the various elements, the endpoint is difficultly controlled by conventional method. Because the BP neural network may realize the free linearity or the non-linear function mapping, it may satisfy the request of the forecast for the endpoint oxygen content. Three kinds of predictive models were developed according to difference of the input variable. Forecast effect is best by taking the endpoint content of carbon, manganese, phosphorus, sulphur and temperature as the input variable. When error scope of endpoint oxygen content is [O] %±0.0100, The off-line prediction hit probability is 75%.