Modeling for prediction steel harden-ability based on IGA-KPLS

Based on the Kernel partial least square (KPLS), a modeling method for the prediction steel quenching degree is proposed in this paper. In order to eliminate the correlations among production parameters, the KPLS is used for feature extraction and Immune genetic algorithm is introduced to optimize the model. With the help of KPLS regression method, a model is then built for the prediction of steel quenching degree. An application study is carried out on the real production data acquired from a steel-making plant. Compared with the existing multiple regression analysis, neural network methods and the ordinary LS-SVM modeling methods, the experimental result shows that the prediction-hit-ratio of the presented method is greatly improved and the precise modeling effect is obtained.