Transfer learning-based deep CNN model for multiple faults detection in SCIM

Deep learning-based fault detection approach for squirrel cage induction motors (SCIMs) fault detection can provide a reliable solution to the industries. This paper encapsulates the idea of transfer learning-based knowledge transfer approach and deep convolutional neural network (dCNN) to develop a novel fault detection framework for multiple and simultaneous fault detection in SCIM. In comparison with the existing techniques, transfer learning-based deep CNN (TL-dCNN) method facilitates faster training and higher accuracy. The current signals acquired with the help of hall sensors and converted to an image for input to the TL-dCNN model. This approach provides autonomous learning of features and decision-making with minimum human intervention. The developed method is also compared to the existing state-of-the-art techniques, and it outperforms them and has an accuracy of 99.40%. The dataset for the TL-dCNN model is generated from the experimental setup and programming is done in python with the help of Keras and TensorFlow packages.