Transfer learning based bearing and rotor fault diagnosis of induction motor

The presented paper investigated the health condition of induction motor through images in different conditions such as healthy, bearing fault, and rotor fault. Novel research is proposed based on a new mixture of the data fusion technique, continuous wavelet transform (CWT), and Deep neural network method in conjunction with transfer learning. The performance of the proposed method was investigated using the data collection of experiments with different challenges such as healthy, bearing fault, broken rotor bar fault under different conditions acquired using National Instruments-based Acquisition. Many classification scores were calculated for different training/testing sets to test the proposed fault recognition system. The problem of manual feature extraction using state-of-the-art shallow learning techniques such as Decision Tree Classifier, Discriminant Classifier, Nearest N eighbor Classifier, and Support Vector Machine Classifier has been solved using the Convolution Neural Network-Based Deep Learning Technique. Furthermore, as CNN develops models from scratch and requires a considerable period, a new transfer learning-based Convolution Neural Network (TL-CNN) method is proposed. This updated method outperforms the shallow learning techniques and CNN by achieving an accuracy of 99.5% using current signatures.