High-sensitivity dual-band microstrip sensor for oil-water mixture analysis with RBF neural network optimization.

Area of Science:

• Electromagnetics and Microwave Engineering
• Sensor Technology
• Machine Learning Applications

Background:

• Accurate characterization of oil-water mixtures is crucial for various industrial and environmental applications.
• Existing methods for dielectric analysis of fluid mixtures often lack sensitivity, selectivity, or compactness.
• Developing novel sensing platforms for real-time fluid composition monitoring remains an active research area.

Purpose of the Study:

• To design, fabricate, and experimentally validate a compact dual-band microstrip sensor for oil-water mixture characterization.
• To investigate the sensor's performance at two distinct resonant frequencies (1.2 GHz and 14.92 GHz).
• To develop a machine learning model for accurate prediction of water concentration based on sensor responses.

Main Methods:

• Fabrication of a microstrip sensor on an FR-4 substrate with dimensions 10.94 × 14.92 mm².
• Experimental measurements of oil-water mixtures with varying concentrations (0-100% in 5% increments).
• Analysis of S-parameter responses and resonance shifts using a Radial Basis Function (RBF) neural network.

Main Results:

• The dual-band sensor achieved high sensitivity (73.5 MHz/εᵣ at 1.2 GHz, 101.48 MHz/εᵣ at 14.92 GHz).
• The RBF neural network model predicted water concentration with R² > 0.99, MSE of 3.24 (%²), and MRE of 3.6%.
• The sensor demonstrated superior performance compared to existing designs in the literature.

Conclusions:

• The proposed compact dual-band microstrip sensor offers a reliable and intelligent platform for dielectric-based fluid analysis.
• The integration of machine learning enhances the accuracy and robustness of oil-water mixture characterization.
• The sensor shows strong potential for real-time industrial and environmental monitoring applications.