Chaotic time series analysis and SVM prediction of alumina silicon slag composition

Alumina silica slag is one of the main raw materials prepared for raw material slurry. Its prediction results by traditional method are poor because of the strong component fluctuation and the long time delay of measurement, which affects the implementation effects of optimal blending system so as to cause instability of the raw slurry´s quality. In the paper, G-P algorithm and the Wolf algorithm of a small amount of data are adopted to analyze chaotic characteristics of silicon slag composition time series. Then, rough set theory is used to construct the generalized phase space of silica slag multi-component time series. Finally, Supporting Vector Machine (SVM) is applied to describing relationship between input and output variables in the generalized phase-space and achieving the exact prediction of alumina silicon slag composition which is beneficial to blending optimization. The experimental results verify the correctness and effectiveness of the proposed method.