Machine Learning methods to estimate odour intensity

Abstract

Odour is a human perception whose relationship with chemical composition is unknown. Contrary to the olfactometric measurement techniques, senso-instrumental methods provide real-time odour monitoring. The study presents a drone equipped with an electronic nose that generates dynamic sensor signals for the classification and quantification of odours in wastewater treatment plants. By calibrating predictive models with Machine Learning algorithms, odour/nonodour samples are classified with 93% accuracy, and odour concentration is predicted 95% limits of agreement within a factor of four, in comparison with dynamic olfactometry measurements