Uncertainty Analysis for EMC Simulation Based on Bayesian Optimization

Abstract

Nonintrusive uncertainty analysis methods are widely applied in the field of electromagnetic compatibility (EMC). When selecting deterministic EMC simulation sampling points as the training set, the existing methods are relatively mechanistic, which can seriously affect the computational efficiency of the uncertainty analysis process. In order to address this computational efficiency issue, this article proposes an EMC simulation uncertainty analysis method based on Bayesian optimization algorithm. This method constructs a Gaussian process surrogate model through a more adaptive training set filtering strategy so as to achieve the goal of giving consideration to both computational efficiency and accuracy. Crosstalk prediction of parallel cables, electromagnetic interference of lightning electromagnetic pulse, and strong nonlinear uncertainty analysis are used to quantitatively verify that the proposed uncertainty analysis method is more efficient and accurate than the traditional non-intrusive method, especially when dealing with complex EMC simulation problems with strong nonlinearity.