Fault Detection of Non-Linear Processes Using Kernel Independent Component Analysis

In this paper, a new non-linear process monitoring method based on kernel independent component analysis (KICA) is developed. Its basic idea is to use KICA to extract some dominant independent components capturing non-linearity from normal operating process data and to combine them with statistical process monitoring techniques. The proposed method is applied to the fault detection in the Tennessee Eastman process and is compared with PCA, modified ICA, and KPCA. The proposed approach effectively captures the non-linear relationship in the process variables and showed superior fault detectability compared to other methods while attaining comparable false alarm rates. Dans cet article, on presente une nouvelle methode de suivi de procedes non lineaire reposant sur l'analyse de composantes independantes du noyau (KICA). L'idee de base consiste a utiliser la methode KICA pour extraire certaines composantes independantes dominantes en capturant la non linearite a partir de donnees de procedes en operation normale et de les combiner avec des techniques de suivi de procedes statistiques. La methode proposee est appliquee a la detection des erreurs dans le procede de Tennessee Eastman puis est comparee aux methodes PCA, ICA et KPCA. L'approche proposee capture de maniere efficace la relation non lineaire entre les variables de procedes et montre une detectabilite des erreurs superieure comparativement a d'autres methodes, tout en atteignant des taux de fausse alarme comparables.