In this letter, a reliable procedure for the expedited design optimization of antenna structures by means of trust-region adaptive response scaling (TR-ARS) is proposed. The presented approach exploits two-level electromagnetic (EM) simulation models. A predicted high-fidelity model response is obtained by applying nonlinear frequency and amplitude correction to the low-fidelity model. The surrogate created this way is iteratively rebuilt and optimized within the trust region framework. The utilization of the correlations between the EM models of various fidelities allows for significant reduction of the design optimization cost. The main contributions of the work are twofold: 1) the application of an ARS for antenna optimization (in particular, making it work with coarse-discretization EM models as low-fidelity models); and 2) the integration of an ARS with a TR optimization framework. The operation and performance of the algorithm are demonstrated using two antenna designs optimized for several scenarios. A comparative study reveals computational benefits of the TR-ARS over direct optimization of the high-fidelity EM model. The reliability of the optimization process is further confirmed by an experimental validation of the fabricated antenna prototypes.
Bibliographical noteFunding Information:
Manuscript received March 21, 2018; revised April 30, 2018; accepted May 4, 2018. Date of publication May 7, 2018; date of current version June 4, 2018. This work was supported in part by the Icelandic Centre for Research under Grant 174114051, and in part by the National Science Centre of Poland under Grant 2015/17/B/ST6/01857. (Corresponding author: Slawomir Koziel.) S. Koziel is with the Engineering Optimization and Modeling Center, Reykjavik University, Reykjavik 101, Iceland, and also with the Faculty of Electronics, Telecommunications and Informatics, Gdansk University of Technology, Gdansk 80-233, Poland (e-mail:,email@example.com).
© 2002-2011 IEEE.
- Adaptive response scaling (ARS)
- antenna optimization
- electromagnetic (EM)-driven design
- surrogate modeling
- trust-region (TR) framework
- variable-fidelity simulations