Abstract: In this paper, we propose a bearing fault diagnosis method based on the VGG16 convolutional neural network with transfer learning. Firstly, the original bearing vibration signal data are preprocessed by a signal-to-image conversion, to generate a target bearing data set. Then, the pretrained VGG16 model is deeply trained by the target data set, and the related parameters are fine-tuned. Finally, the iterated VGG16 model is applied for the fault diagnosis. The above method is verified by the bearing data sets obtained from Case Western Reserve University and University of Cincinnati, respectively. The experimental results show that a diagnostic accuracy of nearly 100% is achieved by our method, which indicates that this method has a good application prospect in the field of the bearing fault diagnosis.