Identification of gear mesh signals by kurtosis maximisation and its application to CH46 helicopter gearbox data

Abstract

The detection and diagnosis of gearbox faults is of vital importance for the safe operation of helicopters. This paper presents a new approach in identifying gear mesh signals for early and effective detection of localised gear faults. Using this approach, the gear mesh signal is identified using a nonminimum phase autoregressive (AR) model by maximising the kurtosis of the inverse filter error signal of the model. Sudden changes in the error signal are usually indications of the existence of localised gear faults in the monitored gear. It is demonstrated using the well-regarded CH46 helicopter aft transmission test data that the approach shows great promise for detecting faults in complex gearboxes.