Real time automatic detection of bearing fault in induction machine using kurtogram analysis

Area of Science:

• Mechanical Engineering
• Signal Processing
• Condition Monitoring

Background:

• Traditional spectral analysis is inadequate for non-stationary vibration signals in real-time diagnosis.
• Early detection of bearing faults is crucial for preventing machine failure and ensuring operational efficiency.

Purpose of the Study:

• To propose and evaluate a novel signal processing technique for incipient, real-time bearing fault detection.
• To leverage kurtogram analysis for enhanced characterization of non-stationarities in vibration signals.

Main Methods:

• Utilized kurtogram analysis, a fourth-order spectral analysis tool, for fault detection.
• Investigated resonance signatures across selected frequency bands to extract representative features.
• Validated the technique through experimental tests on bearings under various fault conditions.

Main Results:

• The proposed kurtogram-based technique demonstrated effectiveness in automatic bearing fault detection.
• Successfully identified incipient faults in real-time, outperforming traditional spectral analysis.
• Extracted representative features by analyzing resonance signatures within specific frequency bands.

Conclusions:

• The developed signal processing technique provides an effective method for automatic bearing fault detection.
• This approach offers a robust foundation for integrated induction machine condition monitoring systems.
• Kurtogram analysis is a suitable tool for diagnosing non-stationary vibration signals in real-time applications.