Procedures to improve convergence and efficiency of genetic algorithms: Application to antenna focusing elements

Design and optimization are difficult tasks when intense electromagnetic computation is involved. For instance, techniques such as gradient, Newton or genetic algorithms (GA) can be used. However, they require some iterative procedure in which analysis of the estimated structure must be analyzed. This is usually achieved by full-wave analysis. In the case of electrically large structures, this step is a lengthy process that produces prohibitive computer expenditure. Using simplified model such as geometrical optics may solve the problem but only partially because the method yields approximated solutions. In this paper, guidelines to enhance the convergence by using small population of individuals are given and examples of large gain antennas are presented.