Department of Aeronautics & Astronautics , MIT
Time: Dec.26.2011, 3 P.M－18 P.M
Room 1216，School of Aeronautics and Astronautics, SJTU
Anomaly Detection in Onboard-Recorded Flight Data Using Cluster Analysis
In the past, aviation safety was mainly improved by learning from accidents. The increasing demands for safety reinforce the industry to be more proactive in safety management. Improvements of safety can be further achieved by utilizing data collected during airlines’ routine operations.
Airlines collect massive amounts of flight data through Flight Operational Quality Assurance (FOQA) programs. The data include hundreds to thousands of flight parameters recorded on board by the Quick Access Recorders (QAR) or the Flight Data Recorders (FDR). The current common practice of FOQA analysis is Exceedance Detection, which triggers an alert when a parameter in the watch list exceeds a pre-defined threshold. Unknown issues remain latent as only known safety issues are examined.
A method of evaluating flight data is developed to identify anomalous flights using data-mining techniques. The method does not require pre-defined thresholds of particular parameters, but detects anomalies that are different from the majority by considering the multivariate time series of flight parameters. The detected anomalies could help domain experts to identify latent patterns of emerging risks in flight operations. Initial evaluations indicate that proposed method is a promising way to identify anomalous flights in airlines’ routine operations.
Lishuai Li is a PhD Candidate in the Department of Aeronautics & Astronautics at MIT, working as a research assistant at the International Center for Air Transportation. She obtained a M.Sc. from the Dept. of Aeronautics & Astronautics at MIT in 2009 and a B.Eng. in Aircraft Design and Engineering in 2007 from Fudan University, China. Her current research interests are in air transportation systems, with a focus on safety analysis.