Efficient stochastic EM studies via dimensionality reduction of polynomial-chaos expansions

Random media uncertainties can have a significant impact on the values of electromagnetic-field components and, as a result, they should not be neglected. However, in many cases certain stochastic parameters may feature a low or even negligible contribution in the variability of the outputs of interest. Having this in mind, we first apply a sensitivity-analysis approach in multi-dimensional random problems, whose results are exploited for the dimensionality reduction of polynomial-chaos (PC) representations. In this way, the computational cost of the PC method is reduced, rendering it more efficient, as no significant accuracy loss is observed. Comparison of numerical computations with results from the Monte Carlo and the original PC techniques indicate the validity of the proposed dimensionality reduction.