Weak Defect Identification for Centrifugal Compressor Blade Crack Based on Pressure Sensors and Genetic Algorithm

Area of Science:

• Mechanical Engineering
• Condition Monitoring
• Signal Processing

Background:

• Centrifugal compressor blades are prone to cracks due to vibration and flow excitation.
• Existing non-destructive testing (NDT&E) methods require shutdowns, are costly, and unsuitable for on-line monitoring.
• Effective on-line monitoring and early defect identification are crucial for compressor reliability and safety.

Purpose of the Study:

• To develop an effective on-line method for identifying weak blade crack defects in centrifugal compressors.
• To address the challenge of weak defect signals masked by noise in pressure pulsation data.
• To enable timely detection of blade defects without interrupting compressor operation.

Main Methods:

• Utilized pressure sensors on the casing to collect on-line airflow pressure pulsation signals.
• Applied Continuous Wavelet Transform (CWT) for signal pre-processing.
• Employed bistable stochastic resonance (SR) with Woods-Saxon and Gaussian (WSG) potential for weak feature enhancement.
• Optimized SR parameters using a Genetic Algorithm (GA).

Main Results:

• The proposed method successfully enhanced weak characteristic frequencies related to blade defects in pressure pulsation signals.
• Experimental validation confirmed the method's effectiveness in identifying subtle crack-induced features.
• Strain tests corroborated the accuracy and reliability of the pressure pulsation signal analysis.

Conclusions:

• The integrated approach of CWT, WSG-based SR, and GA provides a robust solution for on-line monitoring of centrifugal compressor blade health.
• This method offers a cost-effective and timely alternative to traditional NDT&E techniques.
• The findings contribute to improved operational safety and reliability in petrochemical facilities.