Comparison of metrics for peaks enhancement in variable speed conditions

The mathematical model of a faulted ball bearings has been extensively investigated in the last decades. So far, it is widely accepted that a faulted bearing is subject to an unwanted slip in working conditions and this leads to a cyclostationary vibration signal. In literature, various metrics have been devised to track the cyclic frequencies of the vibration signal, based on the statistical definition of cyclostationarity. In variable speed applications, the momentum of the mechanical system and dynamics loads play a fundamental role in the resulting amplitude of the vibration signal. A valuable metric for condition monitoring has to enhance the peaks of vibration signal related to a fault even if the dynamic load is not high. In this paper, five metrics are compared on a benchmark data from a variable speed application. The results indicate time-varying kurtosis (i.d. the moment of the fourth order of the data) as a possible winner.