Antoniadis I., Yiakopoulos C., Patargias T.

Frequency domain based signal processing methods have been shown to present a quite effective behavior in the detection of defects, when applied to the analysis of vibration signals, resulting from rolling element bearings. However, these methods typically require some complex and sophisticated analysis, which renders their application cumbersome for applications requiring unskilled personnel or automated fault detection and trending. Parallel, a number of traditional methods exist, such as the root mean square, the crest factor, the kurtosis, the impulse factor and the shape factor, requiring only direct processing in the time domain. Alternatively to these methods, a morphological index for processing vibration signals has been proposed, addressing the issues of how to quantify the shape and the size of the signals directly in the time domain. The morphological closing using a flat structuring element, results to the extraction of a series of impulses corresponding to the impact series, produced by a defective bearing. This series can offer a better visual inspection of the impulsive signal and contain sufficient quantitative information about the repetition period and the intensity of the impacts. Based on this extracted impulse series, the morphological index can be used as a measure of the development of the defect. In this paper, based on a model for the dynamic behavior of defective rolling bearings, the sensitivity of the morphological index is assessed, compared to the previous traditional time domain indices. The morphological index presents a more uniform and reliable behavior in the cases of different excited natural frequencies, growth of the additive noise level and different propagation media. This fact is then further verified in a case from industrial installations, presenting fault trending analysis of bearings under inner race defect.