The proposed work presents a hybrid signal processing - three level deep learning framework leveraging transfer learning to detect incipient stator inter turn faults in induction motor drives. An experimental test rig is developed using a three phase squirrel cage induction motor to emulate various fault levels. This non-invasive methodology solely relies on the stator currents extracted from the developed test rig to detect the stator inter turn faults under its incipient condition under various conditions of load torque and speed. The framework integrates wavelet transform for initial feature extraction and a three stage deep learning architecture, comprising a convolution layer, a ResNet-50 block with pre-trained weights, and a shallow neural network of classification. The proposed strategy achieved a classification accuracy of 94.37% on the testing dataset. The model is also compared with other existing neural network architectures and displayed superior performance for the SITF diagnosis.
Detection of Incipient Stator Inter-Turn Faults in Induction Motor Drives Using Hybrid Deep Learning Architecture
2024-07-31
1744842 byte
Conference paper
Electronic Resource
English
Detection of Inter Turn Short Circuit Faults in Induction Motor Using Artificial Neural Network
DOAJ | 2020
|