IoT-Driven Regression Tree Models for Efficient Microwave Dielectric Material Characterization: Addressing Non-Linear Cavity Sensing

Abstract

Interconnected microwave dielectric sensing nodes have the potential to revolutionize microwave material processing and design, where microwave dielectric materials characterization (MDMC) with high precision and rapid circuit design are crucial. This research presents an Internet of Things (IoT)‑enabled automated MDMC system designed to tackle the non‑linearity challenges in the extended cavity perturbation regime. Utilizing a cylindrical cavity operating in TE₁₁₁ mode at 5 GHz, the proposed MDMC system is extensively trained on a diverse range of materials through numerous full‑wave 3D electromagnetic simulations. The outputs, i.e., relative permittivity and loss tangent, are derived using advanced machine learning models, including Decision Tree (DT) and Ensemble Learning (EL). A comparative analysis that incorporates simulation, measurement, and predicted permittivity values across varying sample dimensions demonstrates the robustness and accuracy of the DT and EL model. This validates the effectiveness of our high‑quality sensor node and sophisticated data processing techniques within an IoT‑centric framework.