Enhanced discrimination of acetone and ethanol gases via nano-structured MOS sensors and machine learning approaches

Abstract Accurate and sensitive detection of acetone and ethanol vapors using a single sensor is challenging due to their overlapping response characteristics. This study introduces a novel electronic nose system leveraging a straightforward array of nano-structured metal-oxide-semiconductor gas sensors and the thermal fingerprint principle for effective gas discrimination. By applying multiple machine learning algorithms, we identified quadratic discriminant analysis as the most effective model, achieving superior classification accuracy. The model’s robustness was confirmed through evaluations with varying train/test split ratios and dimensionality reduction via principal component analysis, ensuring reliable gas classification and concentration estimation. This approach combines simplicity, high selectivity, and scalability, offering a practical and efficient solution for developing intelligent sensor networks for multi-gas detection and environmental monitoring in real-world applications.