Modeling and prediction of aircraft control surface angles using machine learning

Purpose This study aims to predict the roll and trimmable horizontal stabilizer (THS) angles, which are critically important for the aircraft’s flight stability and controllability, using a multivariate modeling approach. Design/methodology/approach In this study, a prediction model analysis was conducted using real flight data collected from the flight data recorder of the commercial aircraft Airbus A319. A total of 16,864 data points were used, with 70% (11,804 data points) allocated for training and 30% (5,060 data points) for testing. Due to their ability to model nonlinear relationships, the decision tree regressor and random forest regressor algorithms were chosen. Findings The prediction models developed using decision tree regressor and random forest regressor effectively predicted the roll and THS angles. The random forest model outperformed the decision tree in terms of accuracy. Both models successfully captured nonlinear relationships, with random forest showing higher robustness. The results suggest that incorporating additional flight parameters could further enhance the model’s performance. Originality/value In this study, the roll and THS angles, which are crucial for the aircraft’s flight stability and controllability, have been selected as output parameters, differing from traditional approaches. Additionally, real flight data obtained from the Airbus A319 flight data recorder have allowed for the differentiation of input-output parameters and modeling processes from other studies.