Selection of informative frequency band in local damage detection in rotating machinery

Abstract

Problem of informative frequency band (IFB) selection in vibration signal processing for local damage detection is discussed. It is proposed to extend the concept of automatic and objective IFB selection proposed by several authors. Till now, kurtosis was preferred as criterion for IFB search. Thus, it is offered to study set of statistics, namely Jarque–Bera, Kolmogorov–Smirnov, Cramer–von Mises, Anderson–Darling, quantile–quantile plot and a method based on the local maxima approach in order to verify their abilities of IFB selection. Also similarities between them are described. It has been proved by simulation and real data analysis that proposed selectors (because they allow us to “select” frequency band) might be equivalent to the spectral kurtosis (SK) in ordinary cases. Moreover, some of the novel selectors are better, because they are less sensitive to incidental spikes that might occur during the signal acquisition process. Proposed selectors might be (as SK) the basis for filter design for informative signal extraction.