ML-Investigations on Goblet shaped Microstrip Antenna with Defected Ground Structure

The development of compact microstrip patch antennas for extensive electromagnetic simulations is becoming complex in design and time-consuming in process. This paper describes a machine learning enabled method to predict the return loss of a goblet-shaped microstrip patch antenna having a defected ground structure. The proposed antenna has -10dB bandwidth from 4-10 GHz and the maximum resonance at 4 and 8.8 GHz.  A large dataset is generated from the antenna design parameters through parametrization. The dataset is employed to train the ML model and then testing to predict return loss for new goblet-shaped patch design. Performance Evaluation is measured based on the predicted return loss with the help of R2, MAE, MSE, RMSE, and MAPE. The proposed method decreases the number of full-wave simulations, and reduces the time spent on iterative antenna design. The proposed method is used to design reliable next-generation communication systems with the help of ML.