Intelligent Eddy-Current Edge Inspection for Automated Quality Assessment and Resource-Efficient Metal Processing

Metal-cutting operations can generate resource losses not only through the kerf itself but also through subsequent reworking, removal of altered edge material, and processing of workpieces that later prove unsuitable. This study develops and experimentally evaluates an automated eddy-current inspection system intended to characterize metal edges immediately after cutting. The system combines a miniature high-frequency eddy-current transducer, three-axis positioning, digital signal acquisition, and software-based processing. A clad D16AT aluminum alloy specimen with edges produced by laser cutting, cold sawing, and hot shearing was scanned. The air-to-metal transition profiles were described by a logistic function, yielding an electromagnetic transition coordinate xc, a transition parameter s, and the coefficient of determination R2. The fitted xc values were 7.30, 8.76, and 8.93 mm for cold-sawn, laser-cut, and hot-sheared edges, respectively; s was 0.64, 0.59, and 0.67 mm, while R2 was 0.961, 0.959, and 0.941. These quantities are interpreted as comparative electromagnetic descriptors and not as direct measurements of heat-affected-zone depth or defect probability. A scenario calculation based on the displacement of the electromagnetic transition relative to the geometric edge gave apparent material-removal indices of 0.192, 1.127, and 1.236 g per 40-mm edge. Under this explicitly model-based scenario, the laser-cut edge was 8.8% lower than the hot-sheared edge. Complementary measurements showed concordant ordering of the electromagnetic descriptors with roughness, burr height, HV0.1, altered-zone depth, conductivity, and removed-layer mass; the apparent and measured masses differed by 0.3–1.1% for this specimen. The results demonstrate that automated eddy-current mapping can differentiate edge states and provide structured data for routing decisions in resource-efficient and zero-defect manufacturing. Independent-specimen replication, fully traceable physical characterization, and production-scale validation are required before the descriptors can be used as acceptance thresholds or as direct estimates of actual waste.