Fault detection and diagnosis of non-linear non-Gaussian dynamic processes using kernel dynamic independent component analysis

This paper proposes a novel approach for dealing with fault detection of multivariate processes, which will be referred to as kernel dynamic independent component analysis (KDICA). The main idea of KDICA is to carry out an independent component analysis in the kernel space of an augmented measurement matrix to extract the dynamic and non-linear characteristics of a non-linear non-Gaussian dynamic process. Furthermore, as a new method of fault diagnosis, a non-linear contribution plot is developed for KDICA. A comparative study on the Tennessee Eastman process is carried out to illustrate the effectiveness of the proposed method. The experimental results show that the proposed method compares favorably with existing methods.