Extraction of Rotating Machine Sources for Fault Diagnostics using Independent Component Analysis

In this paper we introduce the use of independent component analysis for extraction of machine signal from noisy measurements. ts. The presence of interference noise at the machine defect sificationlas- ification cause misclassification. Consequently, we present a flexible method using the independent component analysis with subband analysis sis for extraction of rotating sources. The proposed method utilizes ICA estimation of unifying structure, includes mutual information and time structure. If the signals have no time correlation, their complexity ty may be achieved by their entropies. We introduce an on-linegorithm,al- ithm, such as an alternative of the algorithm that includes mutual information and time-correlation. Simulation study shows that the combined scheme can effectively extract the signal of interest from the instantaneous and delayed mixtures with noise, in comparison with the conventional method and the proposed approaches. The performance ance of the proposed method is evaluated experimentally on aingbear- ng test bed data using the proposed method and the results show to provide more accurate defect analysis with noisy corrupted data over the conventional method.