Machine Learning Assisted Novel Microwave Sensor Design for Dielectric Parameter Characterization of Water–Ethanol Mixture

In this paper, a novel machine learning assisted microwave sensor is introduced for the dielectric parameter characterization of ethanol liquid samples with different ingredient concentrations. The proposed sensor is designed with structural geometry of two intercoupled spiral resonators on Rogers RO4003 substrate in band stop filter configuration. The fabricated prototype has overall physical size of 15 mm
$\times $
20 mm in 1.8 GHz frequency band. The quantitative sensing performance is numerically analyzed and experimentally verified with the result of good agreement in between. One design novelty in the proposed microwave sensor is that the middle part of main sensing unit is empty for a disposable 3D printed cap to be placed into. These results minor amount of liquid samples to be dropped without the requirement of whole microwave sensor circuit to be replaced for each sample measurement. The liquid samples dropped into the middle part of microwave sensor change the stop band frequency, which is the main parameter to be correlated to dielectric parameters of liquid samples. The proposed microwave sensor has technical potential to be utilized as quality benchmarking equipment for ethanol samples in a compact size with low cost, high sensitivity, reusable, easy fabrication, and machine learning assisted operability features.