Three-Objective Antenna Optimization By Means of Kriging Surrogates and Domain Segmentation

In this paper, an optimization framework for multi-objective design of antenna structures is discussed which exploits data-driven surrogates, a multi-objective evolutionary algorithm, response correction techniques for design refinement, as well as generalized domain segmentation. The last mechanism is introduced to constrain the design space region subjected to sampling, which permits reduction of the number of training data samples required for surrogate model identification. The generalized segmentation technique works for any number of design objectives. Here, it is demonstrated using a three-objective case study of a UWB monopole optimized for best in-band reflection, minimum gain variability, and minimum size. The numerical results indicate that segmentation leads to reducing the cost of initial Pareto identification by around 21 percent as compared to the conventional surrogate-assisted approach.