Research on Bearing Fault Identification of Wind Turbine Based on Deep Belief Network

• Published in: 2020 IEEE 4th Conference on Energy Internet and Energy System Integration (EI2) pp. 4076 - 4080
• Main Authors: Cheng, Shucan, Tao, Wenqiang, Zhao, Yanpu
• Format: Conference Proceeding
• Language: English
• Published: IEEE 30.10.2020
• Subjects: Classification algorithms; Deep belief network; Fault diagnosis; Fault identification; Feature extraction; Frequency spectrum; Frequency-domain analysis; Training; Vibrations; Wind turbine; Wind turbines
• DOI: 10.1109/EI250167.2020.9346678

The failure of wind turbine will greatly decrease the economic benefits of the wind farm. It is thus of great significance to locate the fault in time to ensure the safety and reliability of wind power operations. In this paper, the fault recognition of motor bearing based on deep belief network (DBN) is proposed, which has the advantage that it can learn and extract features independently. The frequency-domain spectrum of motor bearing fault vibration signal collected by the sensor is obtained by Fast Fourier Transform (FFT), then the frequency-domain sequence is normalized, and the training and testing samples are obtained according to appropriate proportion. After learning the fault information contained in the training samples by Restricted Boltzmann Machine (RBM) and finely tuning the global parameters by gradient descent algorithm, the final classifier for the test samples is obtained. The case analysis is carried out on the data set of bearing fault simulation test platform. Each diagnosis scheme includes multiple complex fault condition signals, and the diagnostic results have good accuracy.