Abstract: The development history of fault diagnosis research methods for asynchronous motor bearings is described,and the causes and early symptoms of bearing faults are analyzed. The formula of vibration characteristic frequency of asynchronous motor bearings when faults occur at different positions is deduced,and a fault diagnosis method based on neural network and transfer learning is proposed. The bearing data set of Case Western Reserve University was converted into single-channel gray image,and then the pre-processed image data set was used to train and fine-tune the VGG16 model through transfer learning. Finally,the adjusted VGG16 model was verified on the test data set,and the accuracy rate was close to 100% in the fault classification test. Data-driven fault diagnosis has become an important trend. The on-line vibration monitoring system and hierarchical distribution system of offshore oilfield have laid a good data foundation for the fault diagnosis of asynchronous motor. Deep neural network and transfer learning have a good application prospect.

Key words: asynchronous motor, fault diagnosis, bearing, neural network, transfer learning