Fault diagnosis method based on moving window PCA

Moving window PCA have gain more attentions for its adaptive ability in monitoring industrial process. A new contribution chart is proposed for fault diagnosis when an abnormal behavior is indicated by MWPCA. The contributions of individual process variables to the process behavior changes and the PCA model coefficients can be illustrated in a 3-dimensional chart, which provides more information in process monitoring and fault diagnosis. The effectiveness of the proposed method is demonstrated by the application on the data from a real industrial process.