Evolutionary optimization of electromagnetic devices using advanced operators and population structures

Because of their relative simplicity and robustness in solving difficult design problems, evolutionary techniques such as genetic algorithms (GAs) are quickly becoming widespread tools for the synthesis of electromagnetic devices. Unfortunately, the inherent randomness in these techniques often renders them quite sluggish for optimization problems involving heavy computational loads, e.g. the design of microwave filters without tuning screws or frequency selective surfaces (FSSs). Furthermore, evolutionary techniques may fail even for objective functions involving light computational loads if their nonlinearity deceives the evolutionary technique into settling for a local optimum. As evolutionary techniques are applied to optimization problems of ever increasing difficulty, the algorithms need to be modified to obtain a quality solution with realistic CPU resources. We report on our development of a general software toolbox which extends the art of the evolutionary optimization making difficult problems more tractable by using two separate ideas: advanced operators and a communal population structure. We emphasize our work on the evolution strategies and communal structure, primarily by illustrating their application to the problem of iris coupled waveguide filter design.