Inversion process driven by global optimization heuristics coupled with MESSINE, a fast-running forward model

New global optimization heuristics, Simulated Annealing,(SA) and Tabu Search (TS) have been coupled with MESSINE (Models for Electromagnetic Simplified Simulation In Nondestructive Evaluation), a fast-running eddy current model. The heuristics, issued from combinatorial optimization, have been adapted to continuous variable problems. The model used, MESSINE, simulates here the eddy current inspection of materials, with axisymmetric defects. The cost functions to be minimized are constructed as quadratic distances, in the impedance plane, between synthetic and experimental measurements. Results show that, starting from any defect configuration (without any 'a priori' knowledge and without additional long derivative computations), the optimization process leads to a good defect reconstruction.