Airports Terminal Maneuvering Areas (TMA) and Control Traffic Regions (CTR) are characterized by a dense air traffic flow with high complexity. In nominal operations, approach flight path safety management consists in procedures
which guide the aircraft to intercept the final approach axis, and the runway slope with an expected configuration in order to land. Some abnormal flights are observed and considered as Non Compliant when the intermediate and the final leg intercepting conditions do not comply with the prescription of the operational documentation. This kind of trajectories generates difficulties for both crew and Air Traffic Control (ATC) and may induce undesirable events such as Non Stabilized Approaches or ultimate events like Control Flight Into Terrain (CFIT), in the worst cases. There is a real need for atypical flights detection tools in order to improve safety. In this work, a post-operational detection method based on functional principal component analysis and unsupervised learning will be presented and compared to geometric features.