•  
  •  
 

DEVELOPMENT OF A MICROWAVE INTERFEROMETRY BASED CHEMICAL ANALYZER**

Abstract

This project explores the integration of microwave interferometry and machine learning to achieve near-real-time inference in sensing and diagnostics applications. Microwave interferometry, known for its high sensitivity to permittivity changes in materials, is employed to measure subtle interactions between chemicals and the electromagnetic field produced by a custom-designed bandpass filter. The filter operates within a frequency range of 0–6 GHz and is constructed on a Rogers TMM 13i substrate, chosen for its high dielectric constant and low-loss characteristics. The collected microwave data is processed using advanced signal processing techniques and fed into a machine learning model for inference. The machine learning pipeline is trained on a dataset representing a wide range of chemical interactions, enabling robust classification and prediction capabilities. Feature extraction from the interferometric signal focuses on key parameters such as amplitude, phase shift, and resonance frequency deviations, which correlate strongly with changes in the material's interaction with the electromagnetic field. A custom Python-based control system integrates sensor operations, data acquisition, temperature management, and inference. Temperature control ensures stability in the resonator’s performance, minimizing environmental influences and enhancing measurement precision. Real-time control of the experimental environment, including flow dynamics and thermal conditions, ensures high fidelity in measurements and reliable data acquisition. The system is validated through controlled experiments involving various chemical samples, demonstrating its ability to detect and classify interactions with high precision. The project advances the field of smart sensing by combining the physical precision of microwave interferometry with the analytical power of machine learning. Applications range from environmental monitoring and industrial quality control to advanced material characterization. By achieving near-real-time inference, this system sets the stage for efficient, scalable, and automated detection solutions, addressing critical needs in rapid decision-making scenarios.

This document is currently not available here.

Share

COinS