Fault diagnosis of rolling bearing based on parallel 1DCNN

LIU Wei1, SHAN Xue-yin1, LI Shuang-xi1, ZHANG Zhi-hua1, YAO Si-yu2

(1.College of Mechanical and Electrical Engineering, Beijing University of Chemical Technology, Beijing 100029, 

China;

2.College of Mechanical and Electrical Engineering, Shihezi University, Shihezi 832003, China)

Abstract: Rolling bearing was a common component in rotating machinery. Because of its poor working environment, it was very easy to fail.Therefore, a rolling bearing fault diagnosis method based on parallel 1DCNN (one-dimensional convolutional neural network) was proposed. First of all, the vibration signal from rolling bearing was processed and divided into training set and test set. Then, a parallel 1DCNN model composed of two channels was constructed. The two channels were able to obtain the information of vibration signal in the time domain and frequency domain respectively. It was worth noting that a relatively small convolution core was used to extract the time-domain information, while a relatively large convolution core was used to extract the frequency-domain information. In addition, the parallel 1DCNN model employed the global maximum pooling layer to replace the traditional fully connected layer, which effectively solved the overfitting problem. Finally, the trained parallel 1DCNN model was utilized to process the bearing dataset from Case Western Reserve University.In order to verify the fault diagnosis effect of the parallel 1DCNN model, the model was compared with the traditional CNN model. The experimental results show that the fault diagnosis accuracy of the parallel 1DCNN model is higher than 0.996. Comparing with the traditional single channel CNN model, the parallel 1DCNN model can make full use of the extracted time-domain and frequencydomain feature information, and has better fault diagnosis ability.

Key words: rolling bearing; fault diagnosis;convolutional neural network(CNN);one-dimensional convolutional neural network (1DCNN);deep learning; feature extraction

LIU Wei1, SHAN Xue-yin1, LI Shuang-xi, et al. Fault diagnosis of rolling bearing based on parallel 1DCNN［J］.Journal of Mechanical & Electrical Engineering, 2021,38(12):1572-1578.