Variable-fidelity EM analysis and simplex-anchored regression surrogates for efficient global optimization of...

Area of Science:

• Electrical Engineering
• Computational Electromagnetics
• Microwave Engineering

Background:

• Electromagnetic (EM) simulations are crucial for dependable microwave design but are computationally expensive.
• Global optimization is often necessary for complex microwave circuits (e.g., miniaturized components, metasurfaces) but incurs prohibitive costs.
• Existing methods struggle to balance accuracy and computational efficiency in microwave circuit optimization.

Purpose of the Study:

• To develop a low-cost, efficient, and reliable global optimization procedure for microwave circuits.
• To address the computational expense associated with traditional EM-driven optimization techniques.
• To enable effective optimization for complex microwave designs requiring global search.

Main Methods:

• Utilizing simplex-based regression surrogates to model circuit operating parameters.
• Employing low-resolution EM simulations for the primary global search phase.
• Implementing simplex updating rules to ensure algorithmic convergence.
• Incorporating supplementary fine-tuning with high-resolution EM models for reliability.

Main Results:

• Achieved remarkable cost efficiency through the use of simplified surrogates and low-resolution simulations.
• Demonstrated a perfect success rate in finding satisfactory designs across multiple algorithm runs.
• Reduced the computational cost to approximately sixty high-resolution EM analyses.
• Delivered competitive design quality compared to benchmark optimization methods.

Conclusions:

• The proposed framework offers a highly cost-effective solution for global microwave circuit optimization.
• The combination of surrogate modeling and phased EM simulations ensures both efficiency and reliability.
• This method is particularly advantageous for complex designs where global search is imperative.