Over the course of a long-duration aging of helicopter drivetrain bearings to examine the consumption of grease life, both vibration and acoustic emission sensing was used to monitor the bearing response as the grease life was consumed through this aging. Acoustic emission is evaluated against vibration in terms of signal trends over the course of the experiment. Common signal metrics are calculated to yield condition indicators, and machine learning techniques are applied to the vibration and acoustic emission data. For the 862 hour duration test run equivalent to over 6700 hours on wing, features of these signals trend with increased degree of aging. Autoencoders were used to enrich existing set of traditional condition indicators and principle component analysis was effectively used for feature fusion. This measured trending shows promise for future onboard Health and Usage Monitoring Systems which may adopt new sensing and data analysis modalities to trend the condition of mechanical systems.
展开▼