Uncertainty Quantification in Electromagnetic Computation

In this work, uncertainty quantification is integrated to assess the sensitivity of a finite element (FE)-based numerical model to uncertainties in its input parameters. We compute the sensitivity of wound inductors’ RL parameters extracted from the FE method with respect to geometric and material uncertainties across a wide range of frequencies (from DC to 1 MHz). To that end, we propose a 2D modeling of geometric uncertainties associated with conductors’ positions in the winding window, and we compute the indices associated with a polynomial chaos expansion-based surrogate model of the uncertain FE model using Bayesian inference techniques.