Incipient Fault Diagnosis Based on Moving Window Cumulative Sum Principal Component Analysis and Reconstructed Contribution Plot

In response to the challenge of diagnosing incipient faults in industrial processes, a fault diagnosis method based on moving window cumulative sum principal component analysis (MWSUM-PCA) and reconstructed contribution plot is proposed in this paper. Based on sliding window, the statistics are updated by iterative calculation using real-time data. The average expected difference reconstruction contribution plot algorithm is used for fault identification. The difference between the contribution value of sliding window data and its expected value is used for reconstruction to reduce the impact of noise on the identification results. Through simulation experiments on the Tennessee Eastman (TE) process, and comparison with the results of PCA and K-nearest neighbor (KNN) PCA algorithms, the experimental results comprehensively demonstrate the effectiveness of the proposed method.