Rolling bearing fault diagnosis method

The invention discloses a rolling bearing fault diagnosis method. A multi-task convolutional neural network model is trained, the fault types and the fault severity degrees of rolling bearings are diagnosed through vibration signals of the rolling bearings, and the method comprises the following steps that: S1, measuring vibration signals of nine rolling bearings containing one fault type and onefault severity degree and vibration signals of one fault-free rolling bearing; S2, marking each vibration signal with an 8-bit fault label; S3, establishing a multi-task convolutional neural network model, and training the model by using the data in the S1 and the S2; S4, measuring a vibration signal of a rolling bearing to be diagnosed, and inputting the vibration signal into the multi-task convolutional neural network model to obtain a fault label of the bearing to be diagnosed; and S5, converting the obtained fault label into a fault type and a fault severity degree. The invention aims to provide the simple and eff