Layered classification for improved diagnostic isolation in drivetrain components

This paper investigates the enhancement of current vibration data based health monitoring techniques for helicopter drivetrain components. The authors recently adapted feature processing algorithms and developed a hierarchical classification system to discriminate between multiple bearing failure mechanisms and provide improved maintenance decision aiding. The data was collected from CH-47D/E (Chinook) main swashplate bearings set on a seeded fault test stand. The authors outline the method of using vibration-monitoring data to detect incipient bearing and drivetrain faults and differentiate between rapid and graceful failures. Fault classification was established by applying principle component analysis (PCA) using a hierarchal approach on the feature sets derived from baseline and faulted bearing data at a full range of flight operating conditions. The resulting prognostics and health management (PHM) technique is evaluated as an improvement upon current capabilities to better mitigate the risk of asset loss, reduce life cycle cost induced by unscheduled shutdown, provide fault progression capture, and allow the ability to communicate more actionable information