Antenna Subtractive-Mask Shape Synthesis Incorporating Model Order Reduction

Abstract

Antenna shape synthesis is an advanced design methodology that treats the geometrical shape of radiating structures as free variables, constrained solely by the underlying electromagnetic physics. This approach significantly expands the antenna design space and enables the discovery of high-performance structures that differ from conventional antenna geometries. However, the prohibitive computational cost associated with repeated full-wave simulations during the optimization process has limited the practical adoption of shape synthesis in industry.

This thesis addresses this challenge by introducing model order reduction (MOR) techniques to accelerate the electromagnetic analysis within shape synthesis workflows. Specifically, a novel adaptation of MOR is developed and applied to the electric field integral equation (EFIE) discretized via the Method of Moments (MoM) using the spatial-domain free-space Green's function for perfectly conducting objects. The proposed reduced-order model achieves substantial reductions in computational time compared to traditional MoM solutions, with the efficiency gains becoming particularly pronounced as the number of frequency points increases.

Leveraging this accelerated forward solver, a new subtractive mask-based shape synthesis method is proposed. In this approach, a genetic algorithm systematically identifies and removes groups of basis functions corresponding to geometric regions (masks) on the antenna mesh, while enforcing complex multi-objective performance criteria. The resulting methodology enables the design of unconventional, high-performance antennas in a matter of hours or days rather than weeks, markedly improving the feasibility of shape synthesis for practical applications.

Furthermore, the developed MOR framework is shown to be effective beyond shape synthesis. A novel antenna-circuit co-simulation technique is presented that incorporates the driving beamforming circuitry directly into a single matrix formulation.

The effectiveness of both the reduced-order model and the proposed synthesis and co-simulation techniques is demonstrated through multiple numerical examples, confirming significant improvements in computational efficiency while maintaining high accuracy. This work contributes new tools that enhance the practicality of advanced antenna design methods and opens pathways for the more widespread industrial application of antenna shape synthesis.