Identification of cyclic components in presence of non-Gaussian noise – application to crusher bearings damage detection

In this paper an issue of local damage detection in a rolling element bearing is discussed. The bearing operates in a hummer crusher, thus the vibration signal acquired on the housing contains a lot of impacts that originate in various sources. In the case of local damage detection it is crucial to find a set of cyclic impulses in the signal. These impulses are informative, in spite of impulses related to the crushing process, which are non-informative. In order to find the damage signature we provide feasibility study on a tool based on cyclostationary approach, namely cyclic spectral coherence. This comprehensive analysis includes study on four different signals from bearings in various condition and operating with or without load applied. This analysis is preceded by motivating preliminary analysis where we examine a few widely-used methods for local damage detection.