A method for the outlier flights detection of the final approach based on FOQA data

Aiming at the final approach of the flight which has high risk of fatal accidents, a data-driven method for outlier flights detection is proposed; firstly, three checking points and the performance parameters of final approach are chosen and extracted from the QAR data in FOQA station; secondly, for the outlier detection model training, DBSCAN is proposed when discrete parameters are included in the training datasets, and one-class SVM algorithm is proposed when discrete parameters are not included in the training datasets; finally, the outlier flights deviating from the group characteristics are detected with the trained models at checking point 1000 feet, 500 feet and 300 feet respectively, and the flights present outlier all through the checking points are filtered, which provides important decision support information for the final approach safety management.