Finding hidden spindle bearing defect periods using Ramanujan filter banks

Abstract

Classifying defects is difficult when spindle bearings have multiple faults that interact with each other and result in the measured vibrations exhibiting cyclo- and non-stationarity. For such cases, this paper shows that the Ramanujan sum when used as a basis to construct filter banks helps decompose the signal into its individual components that makes it possible to estimate hidden defect periods. Illustrative examples with different spindle types show how filter banks can identify periods not obvious in the signal’s Fourier spectra, or in spectrograms, or even those left unclassified by the envelope spectrum method or the empirical mode decomposition method.