Rapid global antenna design by simplex regressors and multi-resolution simulations.

Area of Science:

• Electromagnetics
• Antenna Theory
• Computational Intelligence

Background:

• Antenna design optimization is computationally intensive due to high electromagnetic (EM) analysis costs.
• Existing global optimization methods, like nature-inspired algorithms, suffer from poor computational efficiency.
• Surrogate modeling faces challenges with dimensionality and parameter range, limiting accurate metamodel construction.

Purpose of the Study:

• To develop an innovative, computationally efficient algorithm for global parameter adjustment in antenna systems.
• To regularize the objective function by operating in the performance figure space.
• To achieve low computational cost through efficient simulation strategies.

Main Methods:

• A simplex-based search is performed in the space of antenna performance figures (e.g., center frequencies, bandwidth).
• The algorithm employs a simplex updating strategy requiring only one EM analysis per iteration.
• Multi-resolution simulations are utilized, combining coarse-discretization full-wave analysis for global search and medium/high-fidelity simulations for parameter tuning and verification.

Main Results:

• The developed algorithm demonstrates global search capability with remarkably low computational expenses.
• Validation on four microstrip antennas shows an average of around one hundred high-fidelity analyses.
• The proposed method achieves performance competitive with traditional local and global optimizers.

Conclusions:

• The innovative algorithm offers a significant reduction in computational cost for antenna design optimization.
• The simplex-based search in performance space effectively regularizes the objective function.
• This approach provides a viable and efficient alternative for global parameter tuning in antenna systems.