Constrained Space Mapping for Design Optimization of Microwave Circuits
Slawomir Koziel
School of Science and Engineering Reykjavik University, Iceland
Simulation-driven optimization becomes increasingly important in microwave design. Due to the complexity of microwave structures and a growing demand for accuracy, theoretical models can only be used to yield initial designs that need to be further tuned to meet given performance specifications. Also, no systematic design procedures exist for various classes of devices including substrate-integrated circuits or UWB antennas. In such cases, simulation-based optimization seems to be the only reasonable design path. Unfortunately, straightforward approach through employing EM solver directly in an optimization loop is not practical because it is CPU-intensive and often fails. Efficient optimization of microwave structures can be realized using space mapping (SM) [1]-[4]. SM replaces direct optimization of an EM-simulated “fine” model by iterative updating and re-optimization of a surrogate model, a composition of a cheap “coarse” model of the structure under consideration and auxiliary transformations. SM is able to yield a satisfactory design in just a few evaluations of the fine model [1], however, selecting the right combination of SM transformations and their parameters to build a surrogate is not a trivial task, yet critical for performance [5]. Recent results [5], [6] permit some automation of the selection process but they do not guarantee good performance or the algorithm convergence. Here, a novel approach is proposed that exploits over-flexible surrogate model and adaptively adjusted constraints imposed on the parameter extraction and surrogate optimization processes. As a result, improved performance of the SM algorithm is obtained both with respect to its convergence properties and the quality of the final design. Additionally, our algorithm alleviates the surrogate model selection problem.
IEEE Antennas and Propagation Society International Symposium (APSURSI)
2010
July
Due to copyright restrictions articles are not available for download. They are however indexed on this site and search results will therefore correctly indicate if such terms are present in the articles.
