Detection of Broken Rotor Bars in Cage Induction Motors Using Machine Learning Methods

Area of Science:

• Electrical Engineering
• Machine Learning Applications
• Condition Monitoring

Background:

• Squirrel cage induction motors are critical in industrial applications.
• Broken rotor bar (BRB) faults can lead to significant operational issues and failures.
• Effective fault detection methods are essential for reliable motor operation.

Purpose of the Study:

• To evaluate and compare the performance of machine learning methods for BRB fault detection.
• To assess the effectiveness of Decision Tree Classification (DTC), Artificial Neural Network (ANN), and Deep Learning (DL) models.
• To determine the most suitable method for detecting BRB faults under various load conditions.

Main Methods:

• Experimental data collection of line-current signatures from motors with BRB faults.
• Feature extraction using Discrete Fourier Transform (DFT) to analyze the frequency spectrum.
• Development and application of DTC, ANN, and DL models for fault classification.
• Validation using a confusion matrix with metrics like accuracy, precision, recall, and F1-scores.

Main Results:

• DTC demonstrated superior accuracy and precision compared to ANN and DL methods for BRB fault detection.
• DTC exhibited less dependency on motor load conditions, performing well for both unloaded and loaded states.
• Accurate detection of twice-frequency sideband components in stator currents was achieved by DTC.

Conclusions:

• DTC is a highly suitable machine learning candidate for detecting BRB faults in squirrel cage induction motors.
• The method proves effective for motors under varying load conditions using line-current signature analysis.
• DTC offers a robust and accurate solution for predictive maintenance of induction motors.