J. Leoni, A. Palman, M. Tanelli

This paper proposes a post-processing filtering policy for identifying false alarms in a Helicopter Usage Monitoring System (HUMS). HUMS are used to diagnose the status of the helicopters transmission sub-system, which connects the engine to the rotor. However, the high false alarm rate of HUMS can be a limitation in avoiding accidents, despite their ability to significantly reduce accidents. The proposed method relies on timefrequency analysis of vibration signatures for the components of interest and uses one-class support vector machines to learn the healthy behavior distribution. For each alarm triggered by the HUMS, a confidence score is computed according to a weighted Manhattan distance from the center of the healthy distribution. The proposed method was validated on real data from two helicopters monitored for more than three years and was able to filter all false alarms. This method is compatible with existing HUMS and can potentially reduce maintenance costs and increase flight safety.

49th European Rotorcraft Forum 2023, Bückeburg, 2023

Conference Paper

englisch

21,0 x 29,7 cm, 10 Seiten

DGLR-Bericht, 2023, 2023-01, 49th European Rotorcraft Forum 2023 - Proceedings; S.1-10; 2023; Deutsche Gesellschaft für Luft- und Raumfahrt - Lilienthal-Oberth e.V., Bonn

NA

Bestellbar

2023