Bearing fault diagnosis base on multi-scale 2D-CNN model

Bearings play an important role as the connection between the motor and the gear. At present, the data collected by most motor bearing datasets are vibration signals in the one-dimensional time domain, and then one-dimensional convolution or other methods are used to analyze the signals. In this work, a fault diagnosis method based on continuous wavelet transform scalogram (CWTS) and multi-scale convolutional neural network (MS-CNN) is proposed in this paper. Continuous wavelet transform is used to analyze the time-frequency relationship of the signal to extract the frequency information of the one-dimensional signal, and two convolutional neural networks with different kernel sizes are applied to automatically extract different frequency signal characteristics from CWTS data. The experiment has achieved satisfactory results on the CWRU standard dataset.