The Analysis On Blast Furnace Smelting Process And Research On Hot Metal Silicon Content Prediction Model

Blast furnace temperature is an important parameter to judge the running state of blast furnace, and often refers to the temperature of the blast furnace molten iron and slag. The level of temperature directly affects whether the blast furnace smelting process is smooth and the production of hot metal is of high quality. As the blast furnace is a closed container, the furnace temperature cannot be measured directly. During tapping process, the temperature of molten iron will drop, so the measured temperature of molten iron cannot fully characterize the thermal state of the inner part of a furnace hearth. On the contrary, there is no problem of information loss of the silicon content of hot metal, and silicon content is an important index of the quality of molten iron. Therefore, researchers usually take silicon content as chemical temperature and monitoring index of blast furnace temperature condition. This paper uses the silicon content of molten iron as the research object, takes No. 2 blast furnace in Liuzhou Iron and steel as the background, and analyses the source and effect factors of silicon based on mechanism analysis. After singling out input cariables based on mathematical analysis, the prediction model of silicon content is established,and then the characteristics of blast furnace is analyzed in detaile. The result is that the model is more suitable for the blast furnace in actual operating conditions. The specific studies are as follows:(1) Select input variables for the silicon content prediction model based on not only mathematical analysis, but also mechanism analysis. By mechanism analysis of the source of silicon in hot metal and the physical and chemical reactions of silicon material, determine the factors that influence on silicon content in hot metal, and these factors are alternative inputs for the model. After mathematical analysis, select the input variables which have greater relevance, and determine the lag time of the influence of each input variable on silicon content.(2) Considering the uncertainty of the detection time of silicon content in hot metal of blast furnace, the improved unequal interval grey model(IUGM(1,1)) was put forward to fit the silicon content at integral time. The algorithm adds a disturbance factor to unequal interval grey model, and use the silicon content data before and after the integral time to calculate the value of the silicon content, and then the time are matched between input and output variables. Thus the output data is available for subsequent silicon content modeling.(3) Considering the fluctuation characteristics in blast furnace operation, a clustering algorithm is proposed to cluster the samples. Clustering algorithm is used to cluster the samples with similar furnace condition together, and then the silicon content prediction model is builded by the sample of each sub set. At the same time, the fuzzy-C means clustering algorithm is introduced. The setting of the threshold of the fuzzy membership is first proposed, and then the sample which is smaller than the threshold of the fuzzy membership is eliminated. The method makes the samples more compact in each cluster, that is, the furnace condition in each cluster is much more similar. The modeling speed and accuracy are improved by clustering and samples rejection.(4) Establish the prediction model of silicon content in molten iron based on support vector regression(SVR) algorithm with processed input and output data. Support vector regression can effectively avoid overfitting, therefore it is more suitable for industrial complex system modeling than other intelligent algorithms. Finally, using MATLAB to write souse codes, the simulation results verify the validity of the proposed model.