Helicopter Gearbox Mechanical Classification based on Vibration Pattern Recognition

In this paper, unsupervised data analysis methods are used in order to characterize helicopter gearbox mechanical components based on their vibration pattern. Differently from classical vibration health monitoring systems applications, where the objective is to detect early symptoms of impending mechanical degradations, the proposed methodology aims at characterizing the components vibration in their nominal state. The purpose is to identify systematic manufacturing or installation deviations affecting the vibration signature. This allows on one hand for a proactive analysis of the consequences of such deviations, and on the other, it helps explaining vibration indicators variability within traditional health monitoring applications. By combining features extraction, data reduction and clustering techniques, it is shown how it is possible to detect patterns in the vibration signatures from a set of helicopter pinions, ultimately leading to characterizing their main manufacturing differences. These first results give encouraging perspectives for further developments.