Fault Diagnosis of Wind Turbines Bearing based on Particle Filter and Sparse Auto-encoder

In order to improve the fault accuracy and efficiency of fan bearings, a bearing intelligent fault diagnosis method based on the combination of particle filter and sparse encoder is proposed. The vibration signal is preprocessed with particle filter, and then the fault features of the bearing are extracted separately to construct a feature matrix; in order to prevent the network structure from overfitting, the extracted fault features are input into the sparse autoencoder network added to Dropout for fault classification. Finally, the soft-max classifier is used for fault classification and recognition, and compared with the traditional intelligent fault diagnosis method. Experimental results show that, compared with other methods, this method can accurately identify faults.