Extracting Vibration Severity Time Histories from Epicyclic Gearboxes

Abstract

Monitoring epicyclic gearboxes in vital power transition situations is still a challenge. In this paper, we discuss these challenges with long time vibration measurements through two industrial examples. The first are the two gearboxes in the front axle of a load haul dumper (LHD) from Pyhäsalmi mine and the second a two stage gearbox from Kelukoski water power station (WPS). The LHD was monitored almost continuously for nearly two years until its breakdown. The data from WPS was intermittent from a five month period. We discuss how to find stable conditions for comparable measurements in these cases. For this we utilise a tacho signal from the cardan axle of the LHD and power measurements from the WPS. It is found that in both cases second derivatives of acceleration signals, called snap, respond more quickly to changes in vibration severity. In the LHD case we get clear trends for increasing norms of snap signals. The trends are extracted with nonparametric regression. The shorter measurement period of the WPS makes it impossible to say if its changes are only seasonal. Spectral analysis shows increase in high frequency vibration with time in both cases but provides almost no help for detailed diagnostics.