Analysis of flight data

The present invention relates to one kind of analysis by means of an aircraft flight data recorder, the data grouped in the size d of the flight in accordance with a feature vector Xi i flight method at least one of the N flight data recorded during the flight of the aircraft, component of the feature vector corresponding to the data recorded during the flight of the aircraft i, therefore, a feature vector Xi flight defined by i, the method comprising the steps of: for a Gaussian kernel flight characteristics Xi entropy component analysis to obtain normal flight area and their distance with respect to the flight characteristics of the area classification Xi; determining abnormality score for each flight zi i, zi anomaly score Xi with respect to the distance defined by the area of ​​the normal flight flight characteristics; zi abnormality score detecting at least one abnormal flight.