Nonlinear online process monitoring and fault diagnosis of condenser based on kernel PCA plus FDA

摘要

A novel online process monitoring and fault diagnosis method of condenser based on kernel principle component analysis (KPCA) and Fisher discriminant analysis (FDA) is presented. The basic idea of this method is: First map data from the original space into high-dimensional feature space via nonlinear kernel function and then extract optimal feature vector and discriminant vector in feature space and calculate the Euclidean distance between feature vectors to perform process monitoring. Similar degree between the present discriminant vector and optimal discriminant vector of fault in historical dataset is used for diagnosis. The proposed method can effectively capture the nonlinear relationship among process variables. Simulating results of the turbo generator's fault data set prove that the proposed method is effective.