An eye bolt shape slotted microstrip patch antenna design and return loss prediction using machine learning

Abstract

In this study, an eye bolt-shaped microstrip patch antenna (MPA) employing substrate integrated-waveguide (SIW) with and without slot is constructed and assessed. To evaluate the effectiveness of the constructed antenna, simulation, evaluation, and machine learning (ML) methodologies were combined. The design could achieve a gain of 5.5 dB, an optimal bandwidth of 8 GHz (28%), and a minimum return loss of − 38 dB using HFSS software simulations. This antenna's VSWR was 1.02, indicating strong impedance matching, and consequently, a significant efficiency of 87% was attained. In an attempt to forecast the suggested antenna’s return loss, four ML techniques are also used: random forest (RF) regression, extreme gradient boosting (XGB) regression, decision tree regression (DTR), and linear regression. Variance score, mean-absolute error (MAE), root-mean squared error (RMSE), and MSE are employed to assess the four ML model's performance. The XGB regression method outperforms than other three ML techniques in terms of prediction results among the four ML algorithms. Therefore, the proposed antenna is a perfect option for enabling machine-to-machine communications using 5G technology, due to its effectiveness.