Application of Least Squares Support Vector Machines to Predict the Silicon Content in Blast Furnace Hot Metal

Least Square Support Vector Machines (LS-SVM) was applied to predict the silicon content in Blast Furnace (BF) hot metal. Mean square error was calculated using M-fold cross validation and blast volume, coal injection and blast temperature apart from theoretical value and actual value of iron output and blast pressure were selected as input variables. Least square support vector machines belong to a class of kernel methods most common one being Radial Basis Function (RBF). The LS-SVM model was used to do off-line prediction using data collected from blast No.1 at Laiwu Iron and steel Co. An off-line test reveals that model gives excellent predictive accuracy, as low as 10-3 in the magnitude. The on-line application of LS-SVM can give accurate prediction and thus, be helpful to the control of blast furnace iron making.