UHF RFID Microwave Sensor Tag Design for an RSSI-Based Machine Learning Assisted Binary Ethanol–Water Mixture Characterization
Cem Gocen, Merih Palandöken
Abstract
This article proposes a novel microwave sensor-based UHF radio frequency identification (RFID) tag design for the dielectric parameter characterization of binary ethanol–water liquid samples with different concentrations. The sensor tag is designed on an FR-4 substrate with the Impinj M6 Dura chip. The fabricated prototype has an overall physical size of <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$82.1\times19.2$ </tex-math></inline-formula> mm. The mixture samples dropped into the middle part of the microwave sensor tag change the RSSI value from the RFID system, the main parameter to be associated with the water ratio and dielectric parameters in the binary mixture to be used in machine learning (ML). Several ML regression methods have been used for ethanol–water characterization. The GPR model obtained a promising result with <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">${R} ^{{2}}$ </tex-math></inline-formula> = 0.98, RMSE <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$=4.08$ </tex-math></inline-formula> , and MAE = 0.85 for the water ratio in the mixture, <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">${R} ^{{2}}$ </tex-math></inline-formula> = 0.98, RMSE = 3.02, and MAE = 0.66 for the dielectric constant real part, and <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">${R} ^{{2}}$ </tex-math></inline-formula> = 0.99, RMSE = 0.26, and MAE = 0.05 for the dielectric constant imaginary part of the mixture, and the model yielded the highest predictive performance among the four ML models. According to the obtained ML metrics, the water ratio in the binary mixture, the dielectric constant real and imaginary parts of the mixture using the RSSI received over the RFID system, the distance, and the RFID frequency have been successfully characterized. The proposed design has the technical potential to be used as characterization equipment for binary ethanol–water samples with low cost, high precision, and reusable features with ML assistance.