Cloud-based MFL Reference Library for Non-Destructive Flaw Detection in Ferromagnetic Components

This paper presents the development of a simulation-driven, cloud-hosted reference database designed for the detection and localization of subsurface defects in ferromagnetic plates using Magnetic Flus Leakage (MFL) testing with a yoke-type magnetizer. Using FEMM 4.2 magnetostatic models, 500 cases covering irregular, parallelepiped and cylindrical defects were generated at varied positions and depths. Automated LUA scripts produced tangential magnetic flux density (Bt) profiles sampled at 100 points along a line 1 mm above the surface; 450 cases populate an Oracle APEX application and 50 are held out for validation. The application visualizes defect geometry and location and matches new measurements to the library via a distance-based algorithm, returning the top five nearest neighbors to enable rapid, approximate localization or confirmation of defect presence. A sensitivity check - recovering a perturbed in-database profile as rank 1 - illustrates method responsiveness. The system supports both in-situ NDT decision-making and practical training, offering a fast, accessible bridge between simulation and field inspection.