Classification of Testing Maneuvers from Flight Data

A flight data recorder is a device dedicated for recording flight data from a wide range of aircraft sensors. The analysis of the flight data has an important role in the development and testing of avionics. The testing and evaluation of the aircraft responses is often performed using the testing maneuvers. The sensor data from one flight are stored within one flight record, which may contain several testing maneuvers. The goal of this thesis is to identify basic testing maneuvers using measured flight data. The theoretical part describes the flight maneuvers and a related format of the flight data. A research in the field of statistics driven data classification and the probability theory, necessary to understand complex Gaussian Mixture Models, is covered in the analytical part. The thesis also introduces an implementation where the Gaussian Mixture Models are used for the classification of the testing maneuvers in the flight data. The testing of the proposed solution for the data from a flight simulator and a real aircraft showed that the Gaussian Mixture Models provide a suitable solution for this task. A possible future research direction is described in the final chapter.

flight data analysis, classi cation, unsupervised learning, identi cation of flight maneuvers, EM algorithm, Gaussian Mixture Models