Integrated GRM-based efficient multi-performance prediction method for reconfigurable Fabry-Perot antennas.

Area of Science:

• Electromagnetic Engineering
• Computational Electromagnetics
• Antenna Design

Background:

• Reconfigurable antennas require efficient methods to link component states to performance metrics.
• Existing models may lack the accuracy or speed for complex multi-objective optimization.

Purpose of the Study:

• To introduce a novel Generalized Regression Model (GRM) for characterizing antenna performance based on reconfigurable unit activation.
• To develop an integrated framework for both forward prediction and inverse design of reconfigurable antennas.

Main Methods:

• Utilized three parallel Generalized Regression Neural Network (GRNN) subnetworks for modeling return loss, gain, and radiation patterns.
• Optimized GRNN subnetworks using Multiobjective Particle Swarm Optimization (MOPSO) and a recursive correction method.
• Developed an enumeration-based inverse mapping integrated with forward predictions for configuration identification.

Main Results:

• The GRM framework successfully modeled key electromagnetic performance metrics.
• The inverse process effectively identified reconfigurable unit configurations meeting performance constraints.
• Validated on a liquid-based reconfigurable Fabry-Perot antenna, demonstrating significant acceleration in optimization.

Conclusions:

• The proposed GRM provides an effective and efficient approach for optimizing reconfigurable antenna systems.
• This model accelerates the design cycle by enabling rapid exploration of design parameters and performance prediction.