Book: Studies in Applied Electromagnetics and Mechanics 14, Electromagnetic Nondestructive Evaluation (II), Editors: R. Albanese, G. Rubinacci, T. Takagi and S.S. Udpa, IOS Press, Amsterdam (1998), ISBN: 90-5199-375-7, pp.101-106
Multiple Defect Analysis via Neuro-Fuzzy Approaches
Raffaele ALBANESE(1), Francesco Carlo MORABITO(1), Alessandro FORMISANO(2), Raffaele MARTONE(3)
(1) DIEMA, UniversitÓ degli Studi di Reggio Calabria, Loc. Feo di Vito, I-89100, Reggio Calabria, Italy (2) Dipartimento di Ingegneria Industriale, UniversitÓ degli Studi di Cassino, Via De Biase 43, I-03043 Cassino (FR), Italy (3) Dipartimento di Ingegneria dell'Informazione, Seconda Univ. degli Studi di Napoli, Via Roma 29, I-81031 Aversa (CE), Italy
An eddy currents testing system for the identification of a single crack may be
puzzled in the presence of multiple defects. In this paper we propose a hybrid
neuro-fuzzy scheme to deal with the possible presence of two neighboring cracks.
The procedure consists of a fuzzy pre-processor which recognizes the number of
cracks in the region of interest, i.e., classifies the data into the different
classes associated with the number of defects. The second part of the procedure
is based on an array of classical back-propagation neural networks, each of them
specialized for data inversion in each class.