A method of fault diagnosis based on PCA and Bayes classification

By Principal Components Analysis (PCA) method, we can extract the main element from the fault sample set to obtain reduced feature space, which is suitable for fault diagnosis. Bayes method has shown its good classification performance in fault diagnosis, while the real-timing of this method can be guaranteed effectively. By taking advantages of the PCA and Naive Bayes classification, an integrated approach is proposed for the fault diagnosis of chemical process. Firstly, the dimension of industrial data was reduced by PCA method, and the resulting data were discretized to some grades for Bayes classification. The simulation results of TE process show that PCA-Bayes classification is feasible to detect and locate faults quickly with good real time property and high robustness.