Multi-scale autocorrelation via morphological wavelet slices for rolling element bearing fault diagnosis

Fault features of rolling element bearings can be reflected by geometrical structures of the bearing vibration signals. These symptoms, however, often spread over various morphological scales without a known pattern. For this reason, we propose a multi-scale autocorrelation via morphological wavelet slices (MAMWS) approach to detect bearing fault signatures. The vibration measurement of a bearing is decomposed using morphological stationary wavelet with different resolutions of structuring elements. The extracted temporal components are then transformed to form a frequency-domain view of morphological slices by the Fourier transform. Although this three-dimensional representation is more intuitive in terms of fault diagnosis, the existence of the noise may reduce its readability. Hence the autocorrelation function is exploited to produce a multi-scale autocorrelation spectrogram from which the maximal autocorrelation values of all frequencies are aggregated into an ichnographical spectral representation. Accordingly the fault signature is highlighted for easy diagnosis of bearing faults. The effectiveness of the proposed approach has been demonstrated by both the simulation and experimental signal analyses. (C) 2012 Elsevier Ltd. All rights reserved.