Parametric Dataset of AI-Designed Patch Antenna on Multiple Substrates

Abstract

This dataset contains parametric designs and corresponding performance metrics of microstrip patch antennas generated for AI-based modeling and optimization tasks. Each sample represents one antenna configuration defined by its geometrical and material parameters, along with the resulting reflection coefficient S11 in dB at a given operating frequency. The input parameters include feed width and feed length, substrate height and PEC height, patch width and patch length, as well as substrate width and substrate length (all expressed in millimeters). Material information is provided through the substrate type and its relative permittivity (epsilon), while the operating frequency is given in GHz. The dataset covers several dielectric substrates with different permittivities, enabling comparative studies of substrate effects on patch antenna performance. The output variable is the simulated S11 (in dB), which can be used as a target for regression, surrogate modeling, or antenna optimization. This dataset is suitable for researchers working on machine learning for antenna design, surrogate modeling of electromagnetic structures, parametric sensitivity analysis, and design space exploration of microstrip patch antennas. It can be used to train and evaluate models that map patch geometry and material properties to RF performance metrics, or as a benchmark for inverse design and optimization methods.