Fault detection and identification of nonlinear processes based on kernel PCA

A new fault detection and identification method based on kernel principal component analysis (PCA) is described. In the past, numerous PCA-based statistical process monitoring methods have been developed and applied to various chemical processes. However, these previous methods assume that the monitored process is linear, whereas most of the chemical reactions in chemical processes are nonlinear. For such nonlinear systems, PCA-based monitoring has proved inefficient and problematic, prompting the development of several nonlinear PCA methods. In this paper, we propose a new nonlinear PCA-based method that uses kernel functions, and we compare the proposed method with previous methods. A unified fault detection index is developed based on the energy approximation concept. In particular, a new approach to fault identification, which is a challenging problem in nonlinear PCA, is formulated based on a robust reconstruction error calculation. The proposed monitoring method was applied to two simple nonlinear processes and the simulated continuous stirred tank reactor (CSTR) process. The monitoring results confirm that the proposed methodology affords credible fault detection and identification.