The Prediction Research Of Furnace Status As To Tendency To Tcold And Hot Based On Support Vector Machine

Blast Furnace is the upstream steel industry, the main process, as the pillar industry of national economy, an important component of its steel industry development and energy consumption have played an important role. Blast furnace process is a very complex process, and its operating mechanism is often a time delay, nonlinear, characteristic. Blast furnace ironmaking process prediction model, has been automated in the iron problem. Temperature changes in trends can make accurate judgments, will help to improve the foreman of the operation, thus to enhance the blast furnace and coke rate reduction purposes.Support vector machine is derived from statistical theory, developed a kind of machine learning algorithms, it is a good solution to the high dimensionality, small sample, nonlinear, local minimum points, and so the actual problem. Currently, it has been widely used in function approximation, pattern recognition, data mining and other fields. The data is selected from the production daily report based on Baotou 6 # BF (2500m3) from 6 January to 5 July, through calculation of correlation coefficient, analysis of the blast furnace process control variables (wind, air temperature, thermal pressure, pulverized coal injection volume), and state variables (roof pressure of load) and blast furnace hot metal silicon content [Si] (blast furnace hot metal silicon content reflects the chemical heat, can be used to represent the furnace temperature) correlation determine the parameters of the temperature control of the lag time, and on this basis, the establishment of nuclear-based principal component analysis and support vector machines to cool the hot blast forecasting model.This thesis includes the following four aspects: blast furnace, blast furnace expert system overview; blast furnace condition and control parameters of the analysis; kernel principal component analysis and support vector machines; based on KPCA and SVM The blast furnace to cool the heat classification model.The paper used 120 tap numbers as sample numbers, the 60 tap as study sample, the other 60 tap as test sample. To build model, check the model by testing the studied sample. The cool recognize ratio of failure diagnosis is 95%, the hot recognize ratio of failure diagnosis is 100%. In conclusion, the parameter of model is correct and the model has high precision.