Magnetic characterization of steel strips using transient field measurements: global sensitivity analysis and regression from a machine-learning perspective

Abstract

In this contribution, the magnetic characterization of steel strips is studied using synthetic data of field-gradient transients, which have been produced via the finite integration technique. The material law is described and parameterized using the Jiles–Atherton model. The sensitivity of relevant magnetic indicators with respect to the material parameters is then analyzed using two global methods: Sobol’ indices and δ-sensitivity indices. In order to accelerate the evaluation of these quantities, a fast metamodel is built using machine learning techniques from a simulated dataset. The solution of the inverse problem based on a tailored learning framework is tested for the different proposed identifiers, and their suitability for the magnetic characterization of the material in question is finally discussed.