Geometry parameter tuning is an inherent part of antenna design process. While most often performed in a local sense, it still entails considerable computational expenses when carried out at the level of full-wave electromagnetic (EM) simulation models. Moreover, the optimization outcome may be impaired if good initial design is not available. This paper proposes a novel approach to fast and improved-reliability gradient-based optimization of antenna structures. Our approach employs a frequency-based regularization to facilitate relocation of antenna operating parameters to their target values, which increases the chances of identifying a satisfactory design under challenging conditions (e.g., poor-quality starting point). At the same time, computational efficiency of the tuning process is enhanced through the involvement of variable-resolution EM models, and restricting the finite-differentiation sensitivity updates to selected parameters only. The latter are decided upon based on the analysis of the design relocation between the subsequent iterations of the optimization algorithm. The presented technique is validated using three examples of microstrip antennas optimized under different scenarios (matching improvement, gain enhancement, size reduction). The results demonstrate superior performance in terms of reliability and design quality as compared to conventional gradient-based and derivative-free search procedures. At the same time, a significant speedup is achieved over the frequency-regularization-based procedure not using the acceleration mechanisms.
Authors
Additional information
- DOI
- Digital Object Identifier link open in new tab 10.1109/tap.2022.3209281
- Category
- Publikacja w czasopiśmie
- Type
- artykuły w czasopismach
- Language
- angielski
- Publication year
- 2022
