Fault diagnosis method of AC motor rolling bearing based on heterogeneous data fusion of current and infrared image

In order to improve the accuracy of rolling bearing fault diagnosis when the motor is running under non-stationary conditions, an AC motor rolling bearing fault diagnosis method was proposed based on heterogeneous data fusion of current and infrared images. Firstly, VMD was used to decompose the motor current signal and extract the low-frequency component of the bearing fault signal. On this basis, the current signal was transformed into a two-dimensional graph suitable for convolutional neural network, and the data set was classified by convolutional neural network and softmax classifier. Secondly, the infrared image was segmented and the fault features were extracted, so as to calculate the similarity with the infrared image of the fault bearing in the library, and further the sigmod classifier was used to classify the data. Finally, a decision-level fusion method was introduced to fuse the current signal with the infrared image signal diagnosis result according to the weight, and the motor bearing fault diagnosis result was obtained. Through experimental verification, the proposed fault diagnosis method could be used for the fault diagnosis of motor bearing outer ring under the condition of load variation, and the accuracy of fault diagnosis can reach 98.85%.