Analysis of Outdoor Air Quality Using Low-Cost MEMS-Based Electronic Nose and Gas Analyzer With Multivariate Statistical Approach

Abstract

In this letter, MEMS gas sensors-based electronic nose (e-nose) was developed and used for odorant evaluations at different parts of the city. A gas analyzer, in conjunction with sensorial analysis, was performed for different odorous samples. The design of experiments that consisted of eight experimental sets was developed to the selectivity and sensitivity of the developed sensor array following the actual environmental scenario. Advanced multivariate statistical approaches, such as linear discriminant analyses and K-means, were used to describe sample similarity and discrimination ability of the system. The e-nose data processing exhibits satisfactory discrimination between air samples with more than 97% variability. A validated partial least-square (PLS) model foresees good co-relation between e-nose measurement and gas analyzer analysis. Analysis of variance shows that the model is a good fit with significantly reduced RMSE values and high $R^{2}$ values. The finding indicates that an e-nose unit could be a low-cost solution for environmental measurement-based odorant emissions measurement.