Research on Fault Diagnosis of UAV Rotor Motor Bearings Based on WPT-CEEMD-CNN-LSTM
Xianyi Shang, Wei Li, Fang Yuan, Haifeng Zhi, Zhilong Gao, Min Guo, Bo Xin
Abstract
To address the challenge of extracting adaptive fault features for unmanned aerial vehicle (UAV) rotor motor bearings and to meet the high accuracy requirements of bearing fault diagnosis, this paper proposes a neural network-based bearing fault diagnosis method using WPT-CEEMD-CNN-LSTM. Initially, the method applies multiple noise reduction processes to the original vibration signals and enhances their time–frequency resolution through Wavelet Packet Transform (WPT) and Complete Ensemble Empirical Mode Decomposition (CEEMD). This effectively removes noise and generates a high-quality dataset. Subsequently, a Convolutional Neural Network (CNN) is employed to automatically extract deep features, while a Long Short-Term Memory (LSTM) network is used for the time-series modeling, thereby constructing an accurate rotor motor bearing fault diagnosis model. The experimental results demonstrate that the fault diagnosis accuracy of this method reaches 96.67%, which is significantly higher than that of the traditional CNN (85%), LSTM (51.33%), and the CEEMD-CNN-LSTM model with single-signal noise reduction (77.33%). This method also exhibits stronger fault identification and generalization capabilities. This study confirms the effectiveness of combining WPT-CEEMD with CNN-LSTM deep learning techniques for UAV bearing fault diagnosis, providing a high-precision and stable diagnostic solution for UAV health monitoring.