Fault identification for industrial process based on KPCA-SSVM

Abstract

Industrial fault identification is significant for finding fault reason and remedying the potential safety problems. As kernel principal components analysis (KPCA) has excellent performance in nonlinear data processing, a kind of fault identification method is proposed based on KPCA and simple support vector machine (SSVM). KPCA was applied to choose the nonlinear principal component of the model input data space, and SSVM was applied to establish fault identification modeling, which could not only enhance the efficiency of calculation, but also could improve the fault identification ability. The proposed KPCA-SSVM was applied to the Tennessee Eastman Process (TEP). Simulation indicates that this method features high learning speed and good identification ability compared with SVM, PCA-SSVM and KPCA-SVM, and is proved to be an efficient fault identification modeling method. © 2012 Springer-Verlag GmbH.