An Improved Fault Diagnosis Method for Aircraft Motors Based on Electrical and Mechanical Signals

In this paper, a new method for aircraft motors fault diagnosis is proposed, which utilizes Wavelet Transform (WT) and Wavelet Packet Decomposition (WPD) for feature extraction, and Multi-Layer Perceptron (MLP) for aircraft machine fault diagnosis based on the extracted features. This study integrates both electrical and mechanical signals for the purpose of machine fault diagnosis. We initially utilized voltage and current signals from the RR@NTU lab dataset as electrical signals, and vibration signals as mechanical signals, employing these three types of signals as inputs to the model. Finally, we found that current and vibration signals contributed more significantly to the model's accuracy than voltage signals did. Therefore, we finally proposed a fault diagnosis model that combines the energy characteristics of vibration and current signals obtained from WPD with the time-frequency domain characteristics of vibration signals obtained from WT. The experimental results show that the accuracy of machine fault type diagnosis is 97.86% using MLP neural network. This model effectively leverages the characteristics of both electrical and mechanical signals to collaboratively diagnose fault types.