Machine Learning Models for Online Anomaly Detection in Flight Operations
Lucas Coelho e Silva, Mayara Condé Rocha Murça · 2023
View Video Presentation: https://doi.org/10.2514/6.2023-4107.vid Anomaly detection in flight operations data is a prominent approach to delivering actionable information for improved aviation performance as anomalies are often related to critical safety events or inefficient operations. In particular, the online detection of anomalies from streaming data provides the basis for novel real-time monitoring and alerting tools that can lead to more proactive and efficient decision-making during tactical operations. In this paper, we develop machine learning models for the online detection of flight trajectory anomalies from aircraft surveillance data. We focus on a specific type of operationally significant anomaly: go-around maneuvers. Air traffic controllers typically have to react quickly upon the occurrence of a go-around in order to reintegrate the aircraft into the landing sequence and maintain a safe and expeditious traffic flow. We develop a data-driven approach to anticipate such anomalous events for real-time air traffic control decision support. For this, two models are learned and compared in terms of predictive and computational performance: a Gaussian Mixture Model and a Temporal Convolutional Neural Network Model. The models are found to behave similarly, being able to anticipate 67% of the go-arounds with a false positive rate of less than 8%. Moreover, we find that the inclusion of weather features in both models allows for earlier prediction of go-arounds during the approach. In particular, the Gaussian Mixture Model is able to predict 78% of the go-around maneuvers correctly as early as 50 seconds from the runway.