Fault Diagnosis of Rolling Bearing with Imbalanced Small Sample Scenarios

Abstract

Rolling bearing is one of the main components of rotating machinery, timely and accurate fault diagnosis plays an important role in the reliability and safety of modern industrial systems. Under practical working conditions, normal data is abundant and the fault data is rare, the recognition rate of the minority class is low when the neural network is used to deal with these imbalanced datasets. Regarding the above-mentioned problems, a deep convolution fault diagnosis model based on ensemble learning voting method is proposed in this paper. First of all, the one-dimensional vibration signal was segmented through a sliding window for data enhancement. In the second place, the characteristics of the signals were extracted using deep convolutional neural networks. Finally, classification was carried out through the voting method of ensemble learning to realize fault diagnosis. The fault diagnosis models were tested on two different datasets and different imbalance ratios, and the experimental results show that the proposed method can be well applied in imbalanced datasets, which has higher fault recognition accuracy and faster operation.