Fault diagnosis of wind turbine gearbox based on residual neural network

Abstract

Gearbox is a key component of wind turbine drive train, and its failure will lead to equipment downtime, economic loss, even threaten to human life. It is, therefore, necessary to perform fault diagnosis on gearbox. In this article, the fault diagnosis of gearbox is performed based on the residual neural network. In the first place, the one-dimensional vibration signal is used as the input signal data, and a residual neural network is applied to extract features for fault diagnosis. A convolutional neural network model is established for comparison. The feasibility of the proposed method and the superiority of the residual neural network are verified by comparing the number of iterations of the two models with respect to the fault identification accuracy, the value of the loss function, and the confusion matrix of the classification results. The results show that the residual neural network model can achieve higher fault identification accuracy.