The 12th CAA Symposium on Fault Detection, Supervision, and Safety for Technical Processes (CAA SAFEPROCESS 2021)

2 / 2021-06-05 12:19:40

Bearing Fault Diagnosis Based on Residual Neural Network

vibration signal, fault diagnosis, feature extraction, residual network

In the field of mechanical fault diagnosis and prediction, the extraction of bearing fault features is a challenging subject, and the bearing vibration signal is the main basis for analyzing its fault characteristics. This paper proposes a method of bearing fault diagnosis based on residual neural network. The empirical mode decomposition algorithm decomposes the signal into different modal functions, and extracts the modal components containing rich fault information, which can enhance the fault characteristics of the bearing signal. The combination of kurtosis and correlation coefficient performs feature selection on modal components, which can effectively remove the influence of interference and extract the main fault features. In order to further improve the accuracy of fault diagnosis, this paper applies continuous wavelet transform to the enhanced bearing fault signal, and converts the 1-D time-domain signal into a 2-D time-frequency image. The time-frequency diagram is used as the input of the residual neural network, and the residual neural network is used to classify and identify the damage degree of bearing faults. Taking the experimental vibration signal of the bearing as an example, and comparing it with the traditional convolutional neural network and Googlenet network, it verifies the effectiveness and accuracy of the feature extraction method in this paper. The residual network has better convergence and classification performance.