Feature compensation and network reconstruction imaging with high-order helical modes in cylindrical waveguides

Area of Science:

• Non-destructive testing
• Ultrasonic guided wave technology
• Deep learning applications

Background:

• Pipe integrity is vital for oil and gas transport.
• Ultrasonic guided waves (UGW) are effective for defect detection.
• Challenges exist in imaging weak defects with limited views.

Purpose of the Study:

• To develop a high-resolution pipe defect imaging method.
• To address limitations of current UGW inversion techniques.
• To enhance defect detection accuracy and robustness.

Main Methods:

• A stepwise inversion method combining feature compensation and deep learning network reconstruction.
• Utilizing high-order helical guided waves for expanded imaging.
• Finite difference method for forward model establishment.
• Pearson correlation coefficient and wall loss estimation accuracy as key metrics.

Main Results:

• Achieved a 0.9669 correlation coefficient and 96.65% maximum wall loss estimation accuracy on test samples.
• Demonstrated robust imaging performance under varying Gaussian noise conditions (5 dB, 3 dB).
• Laboratory experiments confirmed the method's practical feasibility.

Conclusions:

• The proposed deep learning approach enhances UGW-based pipe defect imaging.
• High-order helical waves improve imaging view and resolution.
• The method shows significant potential for non-destructive testing of pipes and similar structures.