Dual-Driven AI for Eddy Current Array-Based Surface Defect Tomography

Accurate defect identification and characterization in electromagnetic nondestructive testing (NDT) are crucial for ensuring structural integrity. This paper introduces a novel electromagnetic eddy current array perception framework that integrates physical and digital AI to enhance NDT. The framework combines a physical electromagnetic neural network (PEMNN) with a physics-aware reinforcement learning (PARL) algorithm, optimizing real-time electromagnetic sensing and improving sensitivity near defects. Additionally, a sensor-informed diffusion model (SIDM) reconstructs high-resolution images for precise defect analysis. Experimental results demonstrate substantial improvements in sensitivity and accuracy, surpassing traditional NDT techniques, and establishing a new benchmark for real-time, high-precision defect tomography in industrial inspection and diagnostics.