Improved Performance of Fault Detection Based on Selection of the Optimal Number of Principal Components

Abstract

This paper presents a new method for selecting the number of principal components (PCs) in fault detection based on principal component analysis (PCA). On the basis of the proposed fault signal-to-noise ratio (SNR), the optimal number of PCs can be determined. SNR indicates the relationship between the sensitivity of fault detection and the number of PCs. By maximizing the fault SNR, the optimal number of PCs can be selected and the performance of fault detection can be improved. This method is applied to Tennessee Eastman process (TEP) and compared with the cumulative percent variance (CPV) method. The simulation results demonstrate its good Performance.