Machine Learning-Based Microwave Techniques for Dielectric Material Classification

This paper presents two innovative Microwave Non-Destructive Testing and Evaluation (MNDT&E) techniques designed specifically for characterizing planar dielectric materials, regardless of their thickness. These techniques involve measuring the reflection coefficient parameters
$S_{11}$
of the materials using two separate microwave characterization instruments: a monostatic free-space radar and an open-ended rectangular waveguide (OERW). Our objective is to develop a compact, low-power, fast instrument for classifying and evaluating the materials sensed by microwaves across a frequency range varying from 3.95 to 5.85 GHz. These approaches coupled with machine learning (ML) models, are employed and validated within two distinct environmental settings: controlled laboratory conditions and more challenging real-world noisy conditions. Furthermore, a comparative performance analysis is conducted between the two proposed techniques.