Fault Diagnosis Method of Wind Turbine Gearbox Based on Fusion Multispectrogram and Improved CNN Neural Network

Abstract

To solve the problem that the input of single spectrogram cannot fully express the fault feature of the wind turbine gearbox, a fault diagnosis method of the wind turbine gearbox based on the fusion of the multi-sensor spectrogram and the improved CNN neural network is proposed. Firstly, in view of the problem of aliasing of vibration signal components of wind turbine gearboxes, the vibration signals of each sensor are sparsely decomposed to obtain high resonance components including gear harmonic components and low resonance components that may include bearing fault impact components. The high-resonance component and low-resonance component spectrograms of the sensor are fused as the input of the convolutional neural network; secondly, the fault diagnosis model of the wind turbine gearbox that fuses the multispectrogram and the improved CNN neural network is constructed and trained; finally, through QPZZ-II The experimental platform for fault diagnosis of rotating machinery verifies the effectiveness of the proposed method. The results show that the proposed method has a high accuracy of 98.55% for fault diagnosis of wind turbine gearboxes.