Specific Absorption Rate Enhancement using Random Forest Algorithm for THz Applications in Microstrip Antenna

In this paper, specific absorption rate (SAR) of microstrip patch antenna is improved using machine-learning optimization technique. Computation complexity and time consuming design procedures are the issues of traditional antenna design software like ANSYS, HFSS, and CST. The proposed microstrip antenna using machine learning approach exhibits a resonant frequency of 10 THz with a substrate box (60×60×0.6 mm), SAR sphere at the centre (2, 2, 10 mm), tissue phantom box (80×80×40 mm), at the centre (-29.4, -38.2, 1.6 mm) with an area of 29.4×38 mm. The model is designed in a realistic physiological environment with the feed line, ground plane. The designed antenna achieved a reflection coefficient < -29 dB, VSWR of 1, gain of 6.4dBi and an initial SAR value of 2.03 W/Kg which exceeds the prescribed FCC SAR limit. Reduction of 59.38 percent in SAR value is attained by introducing graphene whereas other parameters concerning the performance of the antenna are optimal and fixed. The antenna is optimized using Random Forest model which offers faster and less convoluted design complications with high accuracy and are relevant to 6G communications, autonomous vehicles, ultra-low latency communications and massive IoT networks.