Nonlinear dynamic process monitoring based on lifting wavelets and dynamic kernel PCA

According to dynamic nonlinear and noise properties contained in chemical process data, a combined monitoring approach based on lifting wavelet and dynamic kernel PCA (LWDKPCA) for adaptive monitoring nonlinear process was presented. First, data were de-noised by lifting wavelets, and then the de-noised data were mapped to kernel space, where faults were detected and monitored. To validate the performance and effectiveness of the proposed scheme, LWDKPCA was applied to monitor TE Process. The results showed the method was faster than conventional wavelet and dynamic kernel PCA (WDKPCA). Compared with dynamic kernel PCA (DKPCA), LWDKPCA is better in fault detection and monitoring.