A fuzzy neural approach to localizing holes in conducting plates

In this paper, a fuzzy neural approach is proposed for localizing holes in conducting plates starting from a set of noninvasive eddy current measurements. The underlying ill-posed inverse problem is solved by means of a system that extracts information on the specimen under test from the measurements and implements a priori constraints to facilitate the detection and characterization of the defect. The method utilizes the concepts of fuzzy inference and of fuzzy curves and surface in order to estimate the location of the hole's center from a reduced set of features. A comparison with neural network estimators is carried out.