Advanced techniques for robust gear crack diagnosis under variable speed

Gear root cracks alter both the free and forced responses of the measured signal. While established gear crack diagnostic methods have been based on one of those responses, methods based on free responses have not yet been developed for variable speed applications. This study provides insights into the differences between methods based on free and forced responses and their suitability for practical application. The paper investigates gear root crack diagnosis using three different techniques for both constant and variable speed: a classical approach based on Stewart’s analysis of TSA signals, transfer function removal, and the angle cepstrum comb liftering (ACCL) method. The application of ACCL is extended to transmission error (TE) and displacement signals, which have not previously been investigated. In addition, the paper systematically examines the choice of tine width (TW) in the ACCL method to optimise the signal-to-noise ratio of the ACCL spectrum over this broader application area. These techniques are applied to two crack cases: a single crack with the same number of hunting tooth periods for all situations, and two cracks for which the data have different numbers of hunting tooth periods, to assess the effect of this parameter on the ACCL method. The optimisation of the TW value indicates that TE requires a higher value of TW than the acceleration and displacement signals, even though the overall shape of the displacement spectrum is more similar to that of TE (angular displacement). Using data collected under healthy conditions as the baseline, the comparative results indicate that TE was the strongest discriminator between healthy and cracked conditions. This study has made the following contributions to the research field. First, this is the first paper that systematically documents the differences between forced and free responses, and between additive and multiplicative (i.e. modulation) effects. Second, it advances speed-robust gear diagnostics through a comparative assessment of TF removal and ACCL methods, while extending the application of ACCL to both TE and vibration signals.