Inverse Multi-parameter Identification of Plane Objects Electrophysical Parameters Profiles by Eddy-Current Method.

Abstract

Based on the “accurate” electrodynamic model of Uzal-Cheng-Dodd- Deeds in the interpretation of Theodoulidis, a computer model in the MathCAD software package was created. It allows modeling the measurement process with a surface eddy current probe for objects with continuously changing properties of electrophysical parameters with their piecewise constant approximation. The software product is necessary for the development of a highly productive neural network metamodel which is used to implement the simultaneous inverse identification of the electrical conductivity and magnetic permeability profiles in the process of physical measurements by eddy current probe in real time. The metamodel, which is a model for a model, performs the functions of a carrier of a priori information about the test object obtained at the preliminary preparatory stage as a result of modeling using an “accurate” physical model. This allows us to collect information about the response of the signal of the probe when measuring conditions change, for example, the frequency of excitation of eddy currents in the object, the size of the lift-off between the probe and the conductor surface, varying the profiles of electrophysical parameters, etc. The implementation of the metamodel on neural networks allows us to make it highly productive in the computational sense and use their generalizing properties when constructing it. The computer model was verified by comparing the calculation results for simplemeasurement cases in the COMSOL Multiphysics software package, which confirmed its sufficiently high accuracy and adequacy. A number of computational experiments with its application have been carried out. They have shown the possibility of distinguishing profiles of electrophysical parameters.