Predicting the performance of a graphene-based patch antenna using a machine learning model for terahertz...

Area of Science:

• Electromagnetic Engineering
• Materials Science
• Machine Learning

Background:

• Terahertz (THz) frequency range applications require efficient antenna designs.
• Graphene's unique properties offer potential for novel antenna structures.
• Traditional electromagnetic simulations for antenna characterization can be computationally intensive.

Purpose of the Study:

• To design and simulate a graphene-based microstrip patch antenna for THz applications.
• To develop and evaluate machine learning (ML) models for characterizing antenna performance.
• To assess the potential of ML models as a faster alternative to conventional simulation methods.

Main Methods:

• Antenna design and simulation using CST full microwave studio.
• Development of Artificial Neural Networks (ANN), Random Forest, and Support Vector Machine (SVM) models.
• Training ML models on data from 784 simulations to predict S11, VSWR, gain, efficiencies, and radiation patterns.

Main Results:

• The designed antenna operates in the 1-5 THz range with a maximum gain of 7.5 dBi at 3.2 THz.
• ANN model achieved high prediction accuracy (R-squared of 0.99) with prediction times as low as 0.7 ms.
• ML models effectively captured the nonlinear relationship between antenna geometry and electromagnetic responses.

Conclusions:

• Machine learning models, particularly ANN, offer a computationally efficient and accurate method for characterizing graphene-based THz antennas.
• The proposed ML approach can significantly reduce simulation time compared to traditional electromagnetic methods.
• These findings support the use of ML-driven antenna design for future wireless communication systems in the THz range.