Random model of vibrations for Foreign Object Damage detection in a civil aircraft engine

Aurélien Hazan (1, 2) , Michel Verleysen (1, 2, 3) , Marie Cottrell (1, 2) , Jérôme Lacaille (4)

1 SAMOS - Statistique Appliquée et MOdélisation Stochastique
2 CES - Centre d'économie de la Sorbonne
3 DICE - MachineLearning Group - DICE
4 SNECMA Villaroche [Moissy-Cramayel]

In this article we analyze several vibration time series measured on a real fan test rig before and after it is hit by a flying object. We show first evidence that a windowed autoregressive model may be used to detect the shock after it occurred. We compare these results with a second time series that measures the rotation period of the fan. Lastly we repeat this experiment using adaptive autoregressive estimation instead of the classical AR estimation algorithm. First results show that vibration data allow better prediction. Furthermore, the windowed AR estimation algorithm seems to perform better than the adaptive one.

Mots clés: aeronautics; engine; vibration; time-series; AR

Domaines: Applications [stat.AP]

Aurélien Hazan, Michel Verleysen, Marie Cottrell, Jérôme Lacaille. Random model of vibrations for Foreign Object Damage detection in a civil aircraft engine. Machine Learning for Aerospace International Workshop, Jul 2009, Marseille, France. pp.149-152. ⟨hal-00446258⟩