Title :
Nonlinear Integral Formulation and Neural Network-Based Solution for Reconstruction of Deep Defects With Pulse Eddy Currents
Author :
Preda, Gabriel ; Hantila, Florea I.
Author_Institution :
Electr. Eng. Dept., Politeh. Univ. of Bucharest, Bucharest, Romania
Abstract :
A method for reconstruction of defects buried deep under material surface of conductive nonlinear materials is proposed. Defects are approximated as zero thickness. Simulation of pulse eddy currents is done using an integral FEM method, with a polarization method with over-relaxation to speed up the nonlinear iterations, and a neural network method is used for the reconstruction of defects shape from the simulated signals.
Keywords :
eddy current testing; finite element analysis; iterative methods; neural nets; polarisation; conductive nonlinear materials; deep defect reconstruction; integral FEM method; neural network-based solution; nondestructive testing; nonlinear integral formulation; nonlinear iterations; over-relaxation; polarization method; pulse eddy currents; Coils; Eddy currents; Geometry; Materials; Neural networks; Shape; Training; Eddy currents; neural networks (NNs); nondestructive testing; nonlinear magnetic;
Journal_Title :
Magnetics, IEEE Transactions on
DOI :
10.1109/TMAG.2013.2283194
