Multi-objective optimization of Cassegrain reflector feeds using space mapping surrogate models

摘要

This paper shows how multi-fidelity simulations of symmetrical Cassegrain antennas may be used to rapidly find a representation of the Pareto front for optimization of the feed horn. The method involves finding the Pareto set of non-dominated solutions in a coarse model domain and aligning the fine model responses over the Pareto set by employing a space mapping surrogate model. Two examples are shown; the first uses an analytically defined feed radiation pattern and the second a corrugated horn as feed for the reflector system. For electrically large systems, the method provides an accurate representation of the Pareto front, in the fine model domain, using only a fraction of the fine model evaluations that would be required in a direct optimization, and thus at a significantly reduced computational cost.