The Key Technology Study On FeO Content Of Sinter

In the process of modern iron-making, sinter is the main raw material, the quality of sinter has important effect on blast furnace result. Producing sinter with sinter method not only solves the problem of lean ore iron-making, but improving the metallurgy property of iron - bearing materials, production indices and economic profit of blast furnaces are largely improved. The FeO content is related to the sinter reduction, which is the significant quality guideline of sinter. Predicting FeO content level is important to sinter product.The FeO content is affected by some sintering parameter, the model is complicated, so we only model it by BP neural networks methods. But the converging velocity of BP neural networks methods is slowly, it is difficult to use it to predict the FeO content real time. Sinter end profile image can reflect the FeO content well and real-time, which is hot– spot research.The text analysis and research many cycle of Sinter end profile image, selecting the lacking of Sinter end profile image with external triggering method using before, and raising new method on selecting Sinter end profile image: the best cross section selecting based on the difference method, and use this method select the best cross section image from a series of image and offer the assuring for the next selecting feature parameters of images.The text also make a large improvement on selecting feature parameters, after analyzing the color information of cross section image, selecting different feature parameters by different color component in cross section digital images, as a result ,the selecting processing is largely simplified.The text forecasts the FeO content using fussy C mean clustering method and radial basis function neural network. C mean clustering has the good cluster precision degree, and the RBF network has the characteristic feature of structure simplicity and fast converging velocity. After clustering the feature parameters of sample images, we divide FeO content into 4 classes, aiming at different sample set, using the correspondent sintering parameter to train the RBF function, we get 4 RBF model with smaller design parameter dispersal rate. In the process of forecasting, first using C mean clustering differentiate FeO content degree in this cross section, secondly input the processes parameters of sintering worksite into the correspondent RBF model and calculate, forecast the FeO content.Using the key techniques above and the research foundation before, we develop the sinter FeO content real - time forecasting system by component technique. The system can forecast the FeO content real– time, has a good system stability, simple operation,. The system has been used 1 year in the third sintering plant of a large iron works so far, after comparing forecasting result and testing result, the conforming rate of mean-value is more than 85%, the absolute error is less than±0.3, achieving the expected accuracy.