Fault diagnosis of electro-mechanical actuator based on WPD-STFT time-frequency entropy and PNN

Abstract

Electro-mechanical actuators (EMAs) are increasingly being used as critical actuation devices of the aircraft. It will cause serious accidents once the fault of EMAs occurs, thus the fault diagnosis of EMAs is essential to maintain the normal operation of aircraft. In this paper, a method based on WPD-STFT time-frequency entropy and PNN is proposed to achieve fault diagnosis of EMAs by processing the vibration signals collected by the accelerometer installed in the EMAs. Firstly, the vibration signals are decomposed by wavelet packet to obtain the signal components of different frequency bands, the signal components are subjected to STFT and spectrograms are obtained. Then, time-frequency entropy is calculated and combined with principal component analysis (PCA) for dimension reduction as the feature vector. Finally, the probabilistic neural network (PNN) classifier is introduced to classify the fault modes. The experimental result shows that this method can accomplish the accurate fault diagnosis of EMAs. Moreover, the performance of the proposed WPD-STFT time-frequency entropy method has an advantage over that of WPD-PCA method or STFT combined with mass-moment entropy method for feature extraction.