ANN-Enabled Gain Prediction and Optimization in Dual-Band SIW Antenna Designs for 5G Networks

Artificial neural networks (ANNs) help improve antenna design process by enabling adaptive optimization strategies that address important challenges in 5G wireless systems, including signal interference, limited coverage, and high user density. This study presents an AI-assisted design methodology for a compact dual-band substrate integrated waveguide (SIW) antenna optimized for 5G operation at 28 and 38 GHz. The antenna is implemented on a Rogers RT/Duroid 5880 substrate using a novel slot configuration with strategically positioned vias to enhance radiation characteristics. The fabricated prototype achieves gains of 8.05 dBi at 28 GHz and 7.89 dBi at 38 GHz, with fractional bandwidths of 6.41% (27.491 - 29.277 GHz) and 1.81% (37.496 - 38.179 GHz), while maintaining a return loss below -10 dB across both operating bands. The pivotal contribution of this work is the development of an ANN-based predictive model capable of accurately estimating antenna gain and radiation efficiency from main parameters such as slot dimensions, via size and feedline width. The proposed model demonstrates excellent predictive accuracy, achieving mean squared error values in the range of 0.00 to 0.001 for gain prediction and 0.018 to 0.066 for radiation efficiency estimation. This AI-driven framework significantly reduces design iterations, computational overhead, and prototyping requirements, offering an automated framework for efficient antenna development in next-generation 5G communication networks.