Fault detection of non-Gaussian processes based on modified independent component analysis

In this paper, some drawbacks of both the original independent component analysis (ICA) algorithm and the FastICA algorithm are analyzed as follows: the order of the independent components is difficult to be determined; because of using the Newtonian iteration, FastICA method often leads to local minimum solution, and the suitable source signals are not isolated. To solve these problems, a modified ICA algorithm based on particle swarm optimization (PSO) called PSO-ICA is proposed for the purpose of multivariate statistical process monitoring (MSPM). The basic idea of the approach is to use the PSO-ICA algorithm to extract some dominant independent components from normal operating process data. The order of independent components is determined according to the role of resumption of the original signal. The proposed monitoring method is applied to fault detection and diagnosis in the Tennessee Eastman process. Applications indicate that PSO-ICA effectively captures the independent components.