High-efficacy global optimization of antenna structures by means of simplex-based predictors

Area of Science:

• Electrical Engineering
• Computational Electromagnetics
• Antenna Theory

Background:

• Modern antenna design relies heavily on computationally intensive full-wave electromagnetic (EM) simulations.
• The high cost of EM analysis hinders simulation-driven design procedures like optimization and statistical analysis.
• Global optimization is often necessary but challenging due to the computational demands of nature-inspired algorithms and metamodel limitations.

Purpose of the Study:

• To introduce a novel, computationally efficient procedure for the global optimization of antenna structures.
• To address the limitations of existing methods in terms of computational cost and metamodel reliability.
• To enable robust antenna design through a combination of automated search and gradient-based refinement.

Main Methods:

• A simplex-based automated search is employed, operating on approximated antenna performance figures.
• The method leverages the weakly nonlinear relationship between antenna geometry and performance for computationally inexpensive updates.
• A single EM analysis is performed per iteration, with automated simplex size reduction to ensure convergence, followed by gradient-based refinement.

Main Results:

• The proposed procedure was validated on four microstrip antenna structures.
• Multiple independent runs and statistical analyses confirmed the global search capability of the algorithm.
• The method achieved a low average computational cost of only 120 EM simulations per antenna.

Conclusions:

• The novel optimization procedure demonstrates superior efficacy compared to local optimizers and nature-inspired algorithms.
• The algorithm significantly reduces the computational burden of antenna design, making complex optimization feasible.
• The validated results show satisfactory outcomes across all tested microstrip antenna structures.