An Artificial Intelligence Assisted Distance Variation Robust Microwave Sensor for Biofuel Analysis Applications

Abstract

This letter presents a novel method for distance variation robustness enhancement of microwave resonator-based sensor using artificial intelligence. Since any small change in the distance of the material under the test to the microwave resonator sensors results in significant shifts in their resonance frequency, the performance of these sensors is very susceptible to movements of the measuring system. By utilizing multiple features of the wideband spectrum of the resonators including the frequency, amplitude, and the quality factor of two resonance harmonics of a microwave resonator, a multilayer perceptron (MLP) neural network is trained to measure the volumetric concentrations of biofuel liquids in various liquid to resonator distances. The average errors of as small as 2% for both gasoline and ethanol are measured over a distance variation of as large as from 1 to 6 mm for the liquid under the test from the resonator.