Performance Analysis of Neural Networks for Fault Detection in Induction Motor

Authors:
  • NSN. Sivaraj
    N. Sivaraj
  • BRB. Rajagopal
    B. Rajagopal
Open Access
Publisher website
Google Scholar
Abstract
Condition monitoring and fault identification are essential for preventing damage in industrial machinery, particularly in three-phase squirrel cage Induction Motors (IMs), which are widely used due to their reliability and robust design. This paper compares three advanced techniques for diagnosing Motor faults, including issues like inter-turn faults in the stator, bearing malfunctions and faults from the rotor by analyzing motor current and speed: 1-D Convolutional Neural Networks (1-D CNN), Grey Wolf Optimized Probabilistic Neural Networks (GWOPNN), and Whale Optimized Pattern Recognition Neural Networks (WOAPRNN). The study evaluates each method’s ability to detect and classify faults. Results show that the whale-optimized pattern classification neural network achieves the highest accuracy of 99.15%, making it the most effective method for fault detection. Each technique -offers unique strengths in fault classification and detection, with the goal of enhancing motor reliability and efficiency in industrial environments. By improving fault diagnosis, these methods contribute to reducing downtime, lowering maintenance costs, and increasing the operational lifespan of induction motors.

This publication has 17 references indexed in Scilit: