Machine learning-driven design of wide-angle impedance matching structures for wide-angle scanning arrays

Area of Science:

• Electromagnetics and Antenna Engineering
• Computational Intelligence
• Materials Science

Background:

• Wide-angle impedance matching (WAIM) is crucial for enhancing the scanning range of antenna arrays.
• Traditional WAIM design methods can be computationally intensive and time-consuming.
• Integrating diverse dielectric materials and optimizing complex structures presents significant design challenges.

Purpose of the Study:

• To develop a versatile and efficient design methodology for optimizing WAIM configurations for arbitrary antenna arrays.
• To enhance the scanning range of antenna arrays by incorporating advanced modeling and machine learning techniques.
• To reduce computational resources and design time while improving adaptability to new antenna structures.

Main Methods:

• Modeling a three-layered WAIM structure using generalized scattering matrices (GSMs) with sufficient excited modes for input impedance calculation.
• Integrating machine learning (ML) algorithms, specifically decision tree, bagging, and random forest, for evaluating WAIM characteristics and prediction.
• Employing a genetic algorithm (GA) for efficient determination of optimal WAIM parameters.
• Validating the methodology by designing and testing three matching layers for arrays operating between 9 and 11 GHz.

Main Results:

• The random forest ML model demonstrated superior performance in predicting WAIM behavior, achieving RMSE, R2, and MAPE scores of 0.033, 0.916, and 2.161, respectively.
• Designed WAIMs effectively improved the scanning range of both microstrip and waveguide arrays within the 9-11 GHz frequency range.
• The methodology achieved a calculation time of 0.3 seconds per angle, with a total runtime under one hour and minimal RAM usage (9.7 MB).

Conclusions:

• The proposed design methodology offers an efficient and adaptable framework for optimizing WAIM configurations.
• The integration of ML and GA significantly accelerates the design process and enhances the performance of wide-angle scanning antenna arrays.
• This approach facilitates the development of tools for broader applications of wide-angle scanning arrays.