Composite Fault Diagnosis of Fighter Based on Wavelet Packet Decomposition and Extreme Gradient Boosting Tree

Abstract

This paper studies the diagnosis of composite faults of elevator and pitch angular velocity sensor in F-16 fighter, and proposes a fault diagnosis method based on wavelet packet decomposition and extreme gradient boosting tree (XGBoost). Aiming at the problem that the combined failure of the elevator and pitch angular velocity sensor causes the difference in the time domain of the angle of attack amplitude used for diagnosis to be insignificant, this paper decomposes the angle of attack signal into eight frequency bands through three-layer wavelet packet transform. After decomposition, the energy of each frequency band constitutes a multi-dimensional feature. The high frequency band contains noise and fault information, and its energy distribution can effectively reflect the fault information. Then use the extreme gradient boosting tree to train the fault features to build a multi-classifier. The simulation experiment results verify the effectiveness of the fault diagnosis scheme.