Process Monitoring based on Nonlinear Wavelet Packet Principal Component Analysis

Abstract

For using process operational data to realize process monitoring, kinds of improved Principal Components Analysis (PCA) have been applied to cope with complex industrial processes. In this paper, a novel nonlinear wavelet packet PCA (NLWPPCA) method, which combines input training network with wavelet packet PCA, is proposed. Wavelet packet PCA integrates ability of PCA to de-correlate the variables by extracting a linear relationship with what of wavelet packet analysis to extract auto-correlated measurements. Then methodology of process monitoring based on NLWPPCA is presented. Finally, the proposed approach is successfully applied to two case studies: process monitoring of an eight variables nonlinear process with noise and Tennessee Eastman process.