Fabrication of Fe2O3@SnO2 nanofiber gas sensors to identify VOCs in mixed-gas environments using dynamic modulation and the machine learning algorithm

The cross-interference of multiple gases and their unknown concentrations in mixed environments renders identifying and determining single-phase gas concentrations in mixed environments a challenge for chemical-resistant metal oxide gas sensors. In this study, a dynamic temperature modulation strategy is proposed, and the temperature-controlled sensitivity behavior of porous SnO 2 nanofibers embedded with Fe 2 O 3 hollow spheres (Fe 2 O 3 @SnO 2 ) as a gas sensor under different frequency modulations is explored. Based on differences in their characteristic curves, linear discriminant analysis was employed to extract features and reduce dimensions. Seven different volatile organic compounds were successfully identified and analyzed. Convolutional neural networks were used to identify certain characteristics and predict the concentrations of single-phase gases in mixed-gas environments. Using training and data validation, seven single-phase gases were successfully separated and identified, with an accuracy of 99.8% for n-butanol. Linear regression analysis was also performed on different gas targets, and the actual concentration was highly consistent with the predicted concentration, rendering the Fe 2 O 3 @SnO 2 gas sensors useful for quantitative analysis.