Interpolated Coarse Models for Microwave Design Optimization With Space Mapping
IEEE transactions on microwave theory and techniques
The efficiency of space-mapping optimization depends on the quality of the underlying coarse model, which should be sufficiently close to the fine model and cheap to evaluate. In practice, available coarse models are often cheap, but inaccurate (e.g., a circuit equivalent of the microwave structure) or accurate, but too expensive (e.g., a coarse-mesh model). In either case, the space-mapping optimization process exhibits substantial computational overhead due to the excessive fine model
... fine model evaluations necessary to find a good solution if the coarse model is inaccurate, or due to the cost of the parameter extraction and surrogate optimization sub-problems if the coarse model is too expensive. In this paper, we use an interpolation technique, which allows us to create coarse models that are both accurate and cheap. This overcomes the accuracy/cost dilemma described above, permitting significant reduction of the space-mapping optimization time. Examples verify the performance of our approach. Index Terms-Coarse model, engineering optimization, microwave design, space mapping, space-mapping optimization. I. INTRODUCTION S PACE MAPPING - is a methodology that allows efficient optimization of expensive or "fine" models by means of the iterative optimization and updating of so-called "coarse" models, which are less accurate, but cheaper to evaluate. Provided that the misalignment between the fine and coarse models is not significant, space-mapping-based algorithms typically provide excellent results after only a few evaluations of the fine model. A similar idea is shared by other surrogate-model-based methods -, however, many of them do not use a coarse model: the surrogate model is created by direct approximation of the available fine model data. Space mapping is widely used in the optimization of microwave devices -, -, where fine models are often based on full-wave electromagnetic simulations, whereas coarse models may be physically based circuit models. Recently, space-mapping techniques have been applied to design problems in a growing number of areas (e.g., -). Slawomir Koziel (M'03-SM'07) received the M.Sc. and Ph.D. degrees in electronic engineering from He has authored or coauthored over 100 papers. His research interests include space mapping, circuit theory, analog signal processing, evolutionary computation, and numerical analysis.