Deep Learning Based Bearing Fault Detection Using Multi-Rate Signal Processing for Regulating Shaft Rotational Speed

A multi-rate signal processing method is proposed and combined with a one-dimensional convolutional neural network (1-D CNN) for bearing fault diagnosis. The proposed method regulates data for the shaft rotational speed to generate vibration features for the 1-D CNN network. The proposed method is verified by using a synthetic dataset first; and then a real dataset acquired from our test platform as well as a publicly available dataset to train and validate the 1-D CNN network.