Statistical optimization and neural network-based characterization using a copper-core eddy current probe

Abstract This study presents the design and performance evaluation of a novel eddy-current probe incorporating copper cores to enhance defect detection and characterization in low-conductivity materials used in nuclear applications. The design combines the simplicity of an absolute probe with the directional sensitivity of advanced configurations. The probe generates a highly symmetric magnetic field and a nearly uniform eddy current distribution within the inspected material, thereby improving sensitivity to surface and subsurface defects. Parametric analyses were conducted to investigate the influence of key inspection parameters, including excitation frequency and probe lift-off, on detection performance. To quantify the probe’s capability for defect detection in different materials, a statistical evaluation based on Receiver Operating Characteristic (ROC) curves and the Area Under the Curve (AUC) metric was performed. Furthermore, a multilayer perceptron (MLP) neural network was implemented for data inversion to estimate key crack characteristics, including crack depth, orientation, and the inspected plate’s material type, from simulated impedance signals.