Design of triband circularly polarized hexagon shaped patch antenna using optimized Siamese heterogeneous convolutional neural networks for 5G wireless communication system

Abstract

The advent of 5G wireless communication systems necessitates the development of advanced antenna designs that offer superior performance across multiple frequency bands. Traditional patch antenna design methods, involving iterative simulations, are time-consuming and often insufficient in fully exploring the vast design space and provide less efficiency. To overcome these issues, this work proposes a novel approach for designing a triband circularly polarized hexagon-shaped patch antenna optimized for 5G applications using an Optimized Siamese Heterogeneous Convolutional Neural Network (SHCNN) coupled with a Circle-Inspired Optimization Algorithm (CIOA). Initially, the triband circularly polarized hexagon-shaped patch antenna is designed. The proposed approach leverages SHCNN to learn the relationship between antenna geometry and performance characteristics, utilizing two identical subnetworks with heterogeneous convolutional layers for efficient feature extraction from varied hexagonal antenna geometries. The CIOA, inspired by the properties of circles such as uniformity and symmetry, refines the antenna design suggested by the SHCNN to achieve optimal triband CP performance. This methodology significantly reduces design time by suggesting promising geometries, explores a vast design space for potential novel configurations, and ensures efficient optimization for optimal performance within the desired frequency bands. Applications include compact, high-performance antennas for 5G base stations and user equipment, enhancing multi-band signal transmission and reception. The introduced antenna design is compiled using MATLAB and HFSS platforms. The simulation results of the proposed antenna, employing SHCNNCIOA methods and operating across three frequency bands (triband) such as low (600 MHz - 1 GHz), mid (2.5GHz - 3.7 GHz), and high (24 GHz - 28 GHz), achieve a gain of 8–10 dB, a return loss of less than -20 dB, higher efficiency at 98 %, and a lower VSWR of 1.5 compared with existing designs.