On the Diagnostics of Planet Gear Bearings

Abstract

Bearings play a pivotal role in the rotating machine scenario, due to their ubiquity and importance. A crowd of signal processing procedures have been developed in order to extract information about incipient localised faults in bearings from the measured acceleration signals. In the case of bearings for planetary gear applications, additional complexities are introduced. First, transducers may only be placed on the exterior of the gearbox, usually rather far from bearings. Second, the rotational axes of the planet gears are not fixed, i.e. they move with respect to the gearbox housing and thus to the transducers. As a result, the vibration signature of the planet gear bearings can be altered by the variable transfer path. In this condition, the standard signal processing techniques fail, and the characteristic bearing fault frequencies cannot be determined. On the other hand, global indicators of the bearing health may be used, but they are not able to specify where the fault is located. In this paper, a pre-processing technique is applied to the vibration signals of a planetary gearbox in order to highlight the planet gear bearing signatures. This technique is based on the McFaddens time synchronous averaging method to extract the vibration data relative to each planet. Then, cyclostationary techniques such as the Cyclic Power has be applied to extract the bearing signature.