Metamodel as Input of an Optimization Algorithm for Solving an Inverse Eddy-Current Testing Problem

A method to characterize defects by processing eddy-current testing (ECT) signals is presented. It works in a short time for generic defects and work-pieces. To retrieve the dimensions of a defect, a metamodel-based Particle Swarm Optimization (PSO) is considered. Indeed, using a metamodel as an input of a stochastic method enables to significantly speed it up. The metamodel is generated using an adaptive database generation. Both tube and plate configurations, corresponding with either simulated or laboratory-controlled measured data are considered as illustration. Good accuracy and satisfactory speed of the method are exhibited, additional information besides the inversion results being provided as well, which highlights possibly indeterminate cases, making the method useful also for decision analysis.