Abstract
The paper proposes a method of simultaneous reconstruction of the electrical conductivity and magnetic permeability profiles of planar metal research objects based on the results of single measurements by eddy current probes using surrogate optimization techniques in a reduced compact subspace design and accumulating the full amount of the most important a priori information about the modes of electromagnetic objects. In addition to the information on the response of probe signals to changes in electrophysical parameters, a priori information includes the data on multifrequency sensing and changes in the lift-off between metal research objects and eddy current probes. All the main stages for the implementation of the method of solving the inverse problem are demonstrated, namely, creating a uniform computer quasi-design of the experiment with improved 2D-projections based on LPτ-Sobol’s sequences; creating surrogate models on fully connected deep neural networks; reducing the dimensionality of the full design space using the principal components method of PCA; reconstructing profiles as a result of surrogate optimisation in a compact subspace. Numerical examples of the method are also presented in the paper.