New transfer learning approach based on CNN network for fault diagnosis

In the industry, induction motors operate in difficult environments. Monitoring the performance of motors in such circumstances is significant which can provide a reliable system. This paper intends to develop a new model for fault diagnosis based on the knowledge of transfer learning using the ImageNet dataset. The development of this framework provides a novel technique for the diagnosis of single and multiple induction motor faults. A transfer learning network based on a VGG-19 convolutional neural network (CNN) is implemented, which achieves a high amount of accuracy with minimum training loss compared to the exciting traditional learning methods. Thermal images with different induction motor conditions are captured and applied as inputs to investigate the reliability of the proposed model. The implementation of this task is to use VGG19 (CNN) based pre-trained network with a dense-connected classifier to predict the true class. The use of a transfer learning network allows for the attributes to be automatically extracted and associated with the decision-making part with a quick and faster training time. The experimental results confirm that the proposed model is promising and successfully able to classify the induction motor faults with high classification accuracy of 99.8%. Furthermore, this model can be further used for other applications based on related topics.