Weighted difference principal component analysis for fault detection

“

To improve fault detection performance of principal component analysis (PCA) in nonlinear and multimodal industrial processes, this paper proposes a new fault detection method based on weighted difference principal component analysis (WDPCA). WDPCA first eliminates the multimodal and nonlinear characteristics of the original data using the weighted difference method. Then, PCA is applied to the preprocessed data, neglecting the influences of multimodality and nonlinearity. The WDPCA does not require process knowledge and multimode modeling. The results of simulation examples verify the effectiveness of WDPCA.

”