Prediction of microstrip antenna dimension using optimized auto-metric Graph Neural Network

Abstract

Microstrip antennas are fabricated by a copper-etched printed circuit board (PCB) through an exact patch shapes are arranged at one side and ground planes are arranged at other side. This manuscript proposes a Prediction of Microstrip Antenna Dimension using Auto-Metric Graph Neural Network is optimized with Sheep Flock optimization Algorithm for predicting antenna dimension (MSA-AMGNN-SFOA). Initially, the input datasets are taken through CST microwave studio and Google Sheet is used to records the information. Then, the input antenna parameter is pre-processed by the normalization techniques of Min-Max scaling methodfor feature scaling that enhance the prediction performance.The pre-processed antenna parameters are fed toAuto-Metric Graph Neural Network (AMGNN) for predicting antenna dimension. Auto-Metric Graph Neural Network not exposesany adoption of optimization methods for scaling the optimum parameters andguaranteeingexact prediction. The proposed Sheep Flock Optimization Algorithm (SFOA) is used to optimize the AMGNN weight parameters and it is implemented. The simulation results of the proposed MSA-AMGNN-SFOA design provide higher predicted accuracy and higher than existing methods, such as design of microstrip antenna using GP regression including (MSA-GPR-ANN), Design and optimization of microstrip antenna utilising Gaussian process and support vector machine (MSA-GP-SVM), design of microstrip antenna using deep kernel model including artificial neural network with (MSA-DKL-ANN-PSO) respectively.

Keywords

• Antenna dimension
• Bandwidth
• Microstrip antenna