Compensation of Ambient Variations in a Temperature-Modulated SnO2 Gas Sensor-Based Electronic Nose for Accurate Discrimination of Structurally Similar Volatile Organic Compounds.

The cross-sensitivity and limited selectivity of metal-oxide gas sensors in accurately identifying volatile organic compounds (VOCs) become challenging under climatic interferences, particularly when ambient temperature (T) and relative humidity (RH) fluctuate. This challenge was addressed through a large-scale experimental campaign comprising 972 measurements for training and 567 experiments for validation and repeatability assessments, conducted over seven months using an electronic nose (e-nose). This study constituted the first comprehensive investigation of nine structurally similar VOCs-methanol, ethanol, 1-propanol, 2-propanol, 1-butanol, 2-butanol, iso-butanol, tert-butanol, and acetone-under different T (20-40 °C) and RH (30-50%) conditions, each examined at 12 concentration levels ranging from 100 to 2500 ppm. A thermal shock-induced (TSI) generic SnO2 gas sensor was employed in the e-nose design, enhancing the adsorption-desorption kinetics of oxygen species on the sensing pellet and producing distinct response patterns for each VOC vapor. Accordingly, a hypothetical surface-chemistry model was proposed. By dynamically varying the pellet's operating temperature, this approach mitigated the interfering effects of T and RH on the TSI gas sensor's conductivity. Time-frequency features extracted from the sensor responses were input into multiple classifiers to determine the best-performing and simplest model with practical considerations, achieving an overall accuracy of 95.99% with consistent class-wise performance. Therefore, the proposed e-nose effectively compensated for drift-like effects while reducing production costs and power consumption, making it highly suitable for real-world VOC detection applications.