A Small-Perturbation Automatic-Differentiation Method for Determining Uncertainty in Computational Electromagnetics

Abstract

A new methodology for approximating uncertainties in computational electromagnetics (CEM) solutions is discussed. The proposed method involves the use of a Taylor series expansion to account for small-perturbations (uncertainty) in the input parameters and subsequent use of an automatic-differentiation technique throughout the numerical calculation chain. This approach is denoted as “small-perturbation automatic-differentiation” (SPAD) method and provides certain benefits over other CEM uncertainty methods available in the literature. The method is illustrated in a couple of examples involving a moment-method solution of the electric field integral equation.