Aircraft Control Surface Damage Detection and Classification Using Autoencoder

This paper proposes an algorithm for detection and classifying aircraft control surface damage using an AI model for cause investigation. Control surface damage on fixed-wing aircraft causes structural and aerodynamic changes that affect the flight control system, which was developed using routine flight data; therefore, knowing the type of damage is essential. The proposed algorithm employs AI models for aircraft damage detection (ADD) and damage type classification (DTC) using routine flight and damage occurrence data. The ADD model uses unsupervised learning, whereas the DTC model uses transfer learning, allowing for effective learning even when abnormal data are small. Furthermore, the ADD model generates detection results using the mean absolute error (MAE) and the Mahalanobis distance. In contrast, the DTC model generates the final classification results using the probability accumulation values. The simulation results show that this AI model algorithm can detect control surface failure quickly and correctly identify damage types.