Research and application of ICA technique in fault diagnosis for equipments

Abstract

In order to overcome the difficulty of feature signal extraction from mixed vibration signals, a new method based on Independent Component Analysis (ICA) is proposed to realize separation and filtering for multi-source vibration signals. ICA technique develops from blind source separation (BSS), which solves the problems of information fusion and feature extraction of multi-sensor signals. In this paper, firstly the principle of ICA was briefly introduced and then a good algorithm of independent component analysis, FastICA was presented. Secondly, application in signal separation and filtering with FastICA is studied in fault diagnosis of big petrochemical equipments. Imitation examples and field experiments show that it is feasible to separate and extract feature signal from multi-source vibration signals and indicated that ICA technique is an effective method in signal preprocessing in fault diagnosis of equipments.