An improved automated ultrasonic NDE system by wavelet and neuron networks

Area of Science:

• Non-destructive testing
• Signal processing
• Materials science

Background:

• Digitized ultrasonic testing signals are increasingly used, but their full potential is not yet realized.
• Effective flaw detection relies on distinguishing flaw echoes from microstructural scattering, making signal denoising crucial.
• Smaller defects are harder to detect and prone to false calls, necessitating advanced signal processing techniques.

Purpose of the Study:

• To enhance ultrasonic signal processing for improved defect detection and localization.
• To develop an automated system for defect classification using artificial neuronal networks.
• To integrate numerical modeling for better understanding and prediction of component service life.

Main Methods:

• Wavelet transform was applied for ultrasonic signal denoising and flaw enhancement.
• Artificial Neuronal Networks (ANNs) were employed for automatic defect classification from A-scan data.
• Numerical modeling of ultrasonic wave propagation was performed using transducer characteristics.

Main Results:

• Wavelet transform successfully suppressed noise and improved flaw localization in ultrasonic signals.
• The system demonstrated good defect localization capabilities.
• Numerical modeling provided insights into expected defects and aided in understanding physical phenomena.

Conclusions:

• Wavelet transform is effective for de-noising and enhancing flaw detection in ultrasonic testing.
• Automated defect classification using ANNs can improve reliability.
• Numerical modeling is a valuable tool for predicting component service life and understanding ultrasonic testing outcomes.