Tuning Second Order Cyclostationarity Parameters for the Diagnosis of Rolling Element Bearings

Failures on rolling element bearings usually originate from cracks that are detectable even in their early stage of propagation by properly analyzing vibration signals measured in the proximity of the bearing. Due to micro-slipping in the rollers-races contact, damage-induced vibration signals belong to the family of quasi-periodic signals with a strong 2nd order cyclostationary component. Cyclic coherence and its integrated form are widely considered as the most suitable tools for bearing fault diagnostics and their theoretical bases have been already consolidated. This paper presents how to correctly set the parameters of the cyclostationary analysis tool to be implemented in an automatable algorithm. In the first part of the paper some general guidelines are provided for the specific application. These considerations are further verified, applying cyclostationary tools to data collected in an experimental campaign on a specific test-rig.