Recent signal processing techniques for the detection of localized faults in gearboxes

This paper presents two new techniques for the detection of localized damage in gear trains. These are the Continuous Wavelet Trans-form (CWT) and the Short-Time Beta Kurtosis (STBK). In contrast with the classical spectral analysis, the proposed techniques extract signal features in the time domain and are able not only to detect the presence of a local defect but also pinpoint its exact location along the circumference of the gear under study. The mathematical foundations of the techniques are presented along with the results of applying them to three different test cases: a mathematically perfect gear, an off-the-shelf realistic gear, and lastly one with a cracked tooth. Another novelty is that the techniques are applied in conjunction to both the amplitude and phase modulations of the vibration synchronous average signals. Using both amplitude and phase yields a more positive assessment of both the gear condition and the type of fault at hand.