Application of Neural Networks for the Detection of Rotor Faults in a Squirrel Cage Induction Machine

Abstract

This thesis explores the use of neural network (NN) algorithms for detecting broken rotor bar (BRB) faults in squirrel cage induction machines (SCIMs). BRB faults are common in industrial settings and often lead to significant downtime and maintenance costs. Traditional fault detection methods, such as signal processing and model-based techniques, typically have limitations in their effectiveness. Artificial intelligence techniques, such as NNs, offer a promising solution to improve the detection process. NN algorithms offer a promising alternative, leveraging their ability to learn complex patterns from data. In this study, a dataset of Hilbert envelope signal spectra, obtained using the Fast Fourier Transform (FFT), from SCIMs with various rotor faults was used to train and test the NN algorithm. This algorithm can accurately detect and classify different machine conditions, including the healthy state and BRBs faults. The results demonstrate the effectiveness of neural networks for detecting BRBs faults in SCIMs. The proposed approach provides a practical and efficient solution for early fault detection, enabling timely maintenance and minimizing downtime in industrial applications