ICA-SVM-Based Modeling for Predicting Silicon Content in Blast Furnace Hot Metal

Based on the improved dynamic independent component analysis(ICA) and the support vector machine(SVM),a modeling method for the prediction of silicon content in blast furnace hot metal is proposed in this paper.In order to eliminate the correlations among production parameters,the dynamic ICA is used for feature extraction.With the help of least square SVM which has low computational complexity,a dynamic recursive model is then built for the prediction of silicon content in hot metal,and a genetic algorithm is introduced to optimize the model performance.An application study is carried out on the real production data acquired from a steel-making plant.The experimental result shows that,compared with the existing time series analysis,neural network methods and the ordinary SVM modeling methods,the prediction-hit-ratio of the presented method is greatly improved.