Applications of space mapping optimization techniques to filter design

Abstract

One of the frontiers that remains in the optimization of large engineering systems is the successful application of optimization procedures in problems where direct optimization is not practical. The recent exploitation of surrogates in conjunction with “true” models, the development of artificial neural network approaches to device modeling and the implementation of space mapping are attempts to address this issue. Our original “Space Mapping” concept, first conceived in 1993, and the subsequent Aggressive Space Mapping approach to engineering design optimization will be discussed, along with new variations. Aggressive space mapping optimization closely follows the traditional experience and intuition of designers. It has been amply demonstrated as a very natural and flexible way of systematically optimizing microwave filters. Space mapping optimization intelligently links companion “coarse” and “fine” models of different complexities, e.g., full-wave electromagnetic simulations and empirical circuit-theory based simulations, to accelerate iterative design optimization of engineering structures. New trust region space mapping optimization algorithms will be mentioned. We briefly review the Expanded Space Mapping Design Framework (ESMDF) concept in which we allow preassigned parameters, not used in optimization, to change in some components of the coarse model. Other recent developments include the introduction of the object oriented SMX system to facilitate implementation of our algorithms in conjunction with certain commercial simulators. Extensive filter design examples complement the presentation.