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

This work presents the design, fabrication, and experimental validation of a compact dual-band microstrip sensor for accurate characterization of oil-water mixtures. The sensor, implemented on an FR-4 substrate with overall dimensions of 10.94 × 14.92 mm2, operates at two distinct resonant frequencies, 1.2 GHz and 14.92 GHz, enabling precise dielectric analysis across a wide frequency spectrum. The dual-mode configuration enhances sensitivity, selectivity, and resolution, facilitating the reliable detection of subtle variations in the relative permittivity of test samples. Experimental measurements were performed on oil-water mixtures with purity levels ranging from 0 to 100%, prepared in 5% increments under controlled laboratory conditions to ensure repeatability and accuracy. The resulting resonance shifts and S-parameter responses were analyzed using a Radial Basis Function (RBF) neural network trained to predict water concentration based on extracted microwave features, including resonant frequencies, magnitude responses, quality factors, and phase characteristics. The trained network achieved a coefficient of determination (R2) greater than 0.99, with a mean square error (MSE) of 3.24 (%2) and a mean relative error (MRE) of 3.6%, indicating outstanding predictive precision and robustness. The proposed sensor exhibited sensitivities of 73.5 MHz/εᵣ at 1.2 GHz and 101.48 MHz/εᵣ at 14.92 GHz, demonstrating significantly higher performance compared to other similar designs in the literature. By combining compact geometry, high sensitivity, and machine-learning-based analysis, this work introduces a reliable and intelligent sensing platform with strong potential for real-time industrial and environmental applications requiring accurate dielectric-based fluid composition analysis.