Multiobjective synthesis of electromagnetic devices using nondominated sorting genetic algorithms

Though the basic laws of electromagnetics have been well understood since the time of Maxwell and Hertz, the design of devices based on these laws remains a very complicated procedure. Given an objective function which somehow measures the overall quality of a problem solution, many studies have shown stochastic techniques to be very successful in locating optimal or near-optimal designs in electromagnetic problems. However, most problems in engineering are multifaceted and thus not well treated by such an unwavering charge at a single goal abstracted into an objective function. Instead, tradeoffs must be found which ensure that the device meets all of many possibly interfering design goals. The article proposes a method for finding the set of all optimal tradeoffs (as defined by the concept of Pareto optimality) between conflicting goals inherent in a given design problem. The method, applicable to a plethora of electromagnetic design problems, is demonstrated by application to the design of broadband multilayered microwave absorbers to minimize thickness and reflectance, and to thinned antenna arrays and arrays with digital phase shifters to minimize maximum sidelobe level and beamwidth.