Wind Turbine Rolling Bearing Fault Diagnosis Using t-SNE and GWO-SVM

Abstract

Fault diagnosis of the key component (i. e., bearings) can ensure the normal operation of the wind turbine and reduce the economic loss. In this paper, aiming at the difficulty in bearing feature extraction, a novel fault diagnosis model using refined composite multiscale sample entropy (RCMSE), t-distributed stochastic neighbor embedding (t-SNE), and grey wolf optimization-based support vector machine (GWO-SVM) is proposed. Firstly, RCMSE is introduced to extract the features of bearing signals in different states, and a high-dimensional feature set can be obtained. Then, t-SNE manifold learning is applied to extract the sensitive and low-dimensional feature set. Finally, the RCMSE+t-SNE set is input to the GWO-SVM classifier for classification. The proposed method is applied to analyze the experimental signals of wind turbine bearings, while the results show that it can effectively and accurately identify various bearing faults. Moreover, compared with other methods, the proposed method has higher recognition accuracy.