Recent Advances in Surrogate Modelling of Reflector Antenna Systems

Abstract

Abstract. This paper discusses some of the recent advances in the surrogate based modeling and optimization of reflector antenna systems, and presents several examples. Much of the focus is on the design of reflector systems for radio astronomy applications, where especially tight specifications are placed on some of the important performance metrics. In several cases the response surfaces of these performance metrics are computationally expensive to compute, and, as an added complication, they are often conflicting and need to be optimized in a multiobjective (MO) or Pareto framework. The large numbers of function evaluations required in these optimization loops make direct full wave simulation in the loop intractable, and surrogate based methods may in these cases lead to reliable global MO optimizations. Two main characteristics of reflector antenna responses of interest are exploited, namely their slow (or periodic) variation with design parameters, as well as the availability of a computationally cheap geometric optics approximation. 4302