EMC Uncertainty Simulation Method Based on Improved Kriging Model

Abstract

These days, uncertainty analysis methods have become a hot research topic in the electromagnetic compatibility (EMC) field. The uncertainty analysis method based on the Kriging surrogate model has the unique advantage of not being affected by “dimensional disasters,” and has gradually attracted the attention of researchers. However, the traditional Kriging surrogate model uses a Latin hypercube sampling strategy to select training sets, which is a relatively passive sampling method, and the computational efficiency and accuracy in the practical application process are uncontrollable. This letter proposes an active sampling strategy based on stochastic reduced-order models (SROMs). By improving the fitness function of the genetic algorithm when complete clustering, a new Kriging model is constructed to complete the EMC uncertainty simulation. In the example of parallel cable crosstalk prediction in the published reference, the mean equivalent area method and feature selection verification methods were used to quantitatively evaluate the results, verifying the accuracy improvement of the proposed improvement strategy.