An Adaptive Autogram Approach Based on a CFAR Detector for Incipient Cavitation Detection

Area of Science:

• Mechanical Engineering
• Vibrational Analysis
• Signal Processing

Background:

• Cavitation significantly degrades centrifugal pump performance and lifespan.
• Early detection of incipient cavitation is critical for preventing catastrophic failure.
• Existing envelope demodulation methods struggle with noise and repetitive impacts.

Purpose of the Study:

• To propose an advanced adaptive Autogram approach for incipient cavitation detection in centrifugal pumps.
• To improve signal processing techniques for identifying cavitation-induced cyclostationary features.
• To enhance the reliability of cavitation detection by minimizing false alarms.

Main Methods:

• Development of a cyclic amplitude model (CAM) to characterize cavitation signals.
• Improvement of the Autogram method with character-to-noise ratio (CNR) and CFAR thresholding.
• Integration of CNR for cavitation intensity representation and CFAR for adaptive thresholding.

Main Results:

• The proposed method successfully extracts cavitation features based on the CAM model.
• Achieved a detection rate exceeding 90% for incipient cavitation.
• Maintained a false alarm rate of 5% across various experimental conditions.

Conclusions:

• The adaptive Autogram approach offers a robust solution for incipient cavitation detection.
• This method provides a reliable tool for the predictive maintenance of centrifugal pumps.
• The integration of CNR and CFAR significantly enhances detection accuracy and reduces false alarms.