A novel fault diagnosis method for aircraft actuator based on ensemble model

Abstract

The actuator of aircraft is the direct executor of flight control signal. When it fails, the aircraft will lose control and even crash. However, due to the low frequency of fault occurrence and the lack of real fault data, the amount of health data and fault data is seriously imbalanced. The research on fault diagnosis of aircraft actuator under imbalanced data has practical engineering significance. Inspired by the ensemble model for solving the problem of imbalanced data classification, this paper proposes an innovative ensemble model called ensemble deep auto-encoder based extreme learning machine (DELM-AE). DELM-AE is a deep network constructed by multi-layer extreme learning machine based auto-encoder, which has the advantages of strong feature mining ability, high accuracy and fast speed. Firstly, the fault simulation model of flight control actuator is established, and then residual analysis and feature extraction are carried out on the data. Finally, compared with other common shallow model (extreme learning machine, support vector machine, back propagation neural network), ensemble model (random forest) and deep networks, the advantages of the proposed method in accuracy, processing speed, robustness and ability to deal with imbalanced data are proved.