Differential Alternating Current Field Measurement with Deep Learning for Crack Detection and Evaluation.

Area of Science:

• Materials Science
• Non-destructive Testing
• Artificial Intelligence

Background:

• Crack detection is crucial for structural integrity.
• Conventional methods face challenges with noise and lift-off effects.
• Accurate crack dimension evaluation is often difficult.

Purpose of the Study:

• To introduce a novel differential Tunnel Magnetoresistance (TMR)-based ACFM probe for enhanced crack detection.
• To integrate deep learning for precise crack dimension evaluation.
• To mitigate the lift-off effect and external noise in TMR-ACFM sensing.

Main Methods:

• Fabrication of a miniature differential TMR probe utilizing a Wheatstone bridge configuration.
• Development and training of a Convolutional Neural Network (CNN) with Convolutional Block Attention Module (CBAM) for crack analysis.
• Experimental validation comparing the differential probe with conventional probes.

Main Results:

• The differential TMR probe demonstrated a quality factor improvement exceeding an order of magnitude.
• Signal-to-noise ratio was enhanced by over 3 dB compared to conventional probes.
• The CNN+CBAM network achieved high-precision crack dimension prediction with low relative errors (e.g., 0.201% for length, 0.709% for depth, 7.224% for width).

Conclusions:

• The proposed differential TMR-ACFM probe integrated with deep learning offers a cost-effective and high-performance solution for crack detection.
• The system significantly improves signal quality and enables quantitative evaluation of crack dimensions.
• This approach shows substantial potential for advanced structural health monitoring applications.