Multiscale Convolutional Neural Network for the Fault Diagnosis of Rolling Bearing

Fault diagnosis of rotating machinery is of great significance to ensure the reliable operation of machinery and equipment. During the operation of mechanical equipment, there will be many noise sources that interfere with the vibration signal, and destroy the characteristic value of the vibration signal, which is difficult to fault diagnosis. Therefore, this paper proposes a multiscale convolutional neural network (MS-CNN) fault diagnosis algorithm. The algorithm has convolution kernels of multiple scales, which can extract the characteristics of vibration signals containing noise sources from different scales. Tested with a public data set containing noise, the fault diagnosis rate can be as high as 96% when the signal- to-noise ratio (the ratio of useful signal power S to noise power N) is -10 dB. Compared with the other four algorithms, the results show the superiority of MS-CNN.