par Haarscher, A.;De Doncker, Philippe ;Lautru, David
Référence Progress in Electromagnetics Research M, 21, page (149-161)
Publication Publié, 2012
Article révisé par les pairs
Résumé : Up to now, ray-tracing simulations are commonly used with a deterministic approach. Given the input parameters, the ray-tracing algorithm computes a value for the electric field. In this paper, we present a method that aims at computing the mean and standard deviation of the electric field. More precisely, we aim to obtain the probabilistic content of the electric field value and direction. We assume that this uncertainty results from input random variables which we consider uniformly distributed. Since ray-tracing computations have a high computational cost, we use spectral methods in order to optimize the number of simulations. We consider 2D electromagnetic propagation for the multi-path components, which can interact with the environment through four processes: transmission, single reflection, double reflection and diffraction. These are modelled using adequate coeffcients. In order to calculate the polynomial chaos expansion coeffcients, we use the projection method and Gauss-Legendre quadratures. These coeffcients can then be used to determine the Sobol indices of input parameters. This is done in order to neglect variables in practical computation of the uncertainties.