Detecting anomalous behaviors in multivariate systems using machine learning
Benjamin Farcy, Alric Gaurier · Open Archive Toulouse Archive Ouverte (University of Toulouse) · 2020
Aircrafts are complex systems that are generating more and more data. An Airbus A320 equipped with FOMAX (Flight Operations and MAintenance eXchanger) records 24000 parameters and a Pratt & Whitney PW1000G GTF engine incorporates 5000 sensors, leading to terabytes of data being recorded for each flight. Modernization of military aircrafts and usage of Unmanned Aerial Vehicle (UAV) fleets also lead to large quantity of data. Detecting anomalies in those systems is valuable as a warning system to detect unexpected behaviors, faulty systems to be replaced or even discard readings of faulty sensors.