Modeling and design of high-frequency structures using artificial neural networks and space mapping

The increasing complexity of high-frequency electromagnetic (EM) structures motivates research for efficient modeling approaches. There are two main targets for developing such structures. The first is to use these models to optimize the original time-intensive "fine" EM structure. In other words, a "surrogate" model is used as a substitute of the EM structure. This surrogate model may be iteratively updated to improve its accuracy as more and more EM simulation data become available. The second application is to use this model in actual simulations instead of using the EM simulation. In this case, the model construction is carried out beforehand using a given set of fine model simulations. Two of the main approaches used in surrogate-based modeling and optimization of high-frequency EM structures include artificial neural networks (ANNs) and the space mapping (SM) approach. Both of these approaches have been applied in a wide range of applications for both modeling and optimization of high-frequency EM structures. This chapter provides a brief review of the basics of both approaches and their current applications in the computer-aided design of high-frequency structures. © 2012 Elsevier Inc.