Determining effectiveness based on the accuracy criterion of the optimization method for reconstructing profiles in the PCA space

Abstract

This study presents a method for simultaneously reconstructing the electrical conductivity and magnetic permeability profiles of planar conductive objects using surface eddy current probes, leveraging surrogate optimization with deep neural network metamodels in reduced PCA-dimensional spaces. The approach addresses the computational complexity of inverse problems by incorporating apriori information and applying dimensionality reduction, achieving high accuracy in profile reconstruction as demonstrated in comprehensive numerical experiments. The results confirm the effectiveness and potential of this method for nondestructive testing and material characterization in various industrial applications