Generalized Interval Arithmetic for Electromagnetic Sensitivity Analysis

This work presents a novel approach for sensitivity analysis of electromagnetic (EM) systems. A generalized analytic strategy based on Interval Arithmetic (IA) is proposed to address cases where the input-output relationship between tolerance-affected parameters and system performance is unavailable in closed form. The method begins by defining an analytic surrogate model of the system transfer function using a learning-by-example approach with available input-output data. This model is then extended to intervals using IA, providing inclusive yet finite performance bounds for the system output response when control parameters have unknown but bounded tolerances. The method is validated through a comparison between the proposed approach and a conventional IA-based method for tolerance analysis of a linear array of isotropic antenna elements.