A data generation approach for surrogate models for magneto‐static simulations

Abstract In this paper, we propose an efficient numerical scheme for the prediction of the magnetic stray fields in two‐dimensional random microheterogeneous materials. Since data‐driven models require thousands of training datasets, Finite Element Method simulations appear to be too time consuming. Therefore, a stochastic model based on Brownian motion, which uses an efficient evaluation of stochastic transition matrices, is used as a Poisson solver to generate training data.