Towards a 3D-printed IoT electronic nose for multipurpose aromatic discrimination : an open source, low cost, and portable solution tested on wine.

Abstract

The sense of smell or olfaction is a primary human sensory system and perhaps the most evocative of the five senses eliciting strong emotional and cognitive responses of both attraction and repulsion. The human olfactory system can detect more than 10000 odors and can discern up to 5000 of them. The human ability to discriminate and detect different odors is however very limited, highly subjective and prone to fatigue. As a result, there have been increasingly many attempts at mimicking the sense of smell in what has now been known as an electronic nose or an e-nose. In the biological olfactory system, the neural network plays the crucial role of coding, processing, and classifying the different odors based on the neural signals received from the olfactory receptors’ interaction with odorants. Similarly, in an e-nose, an array of non-specific chemical sensors reacts to the aerosols, altering them to varying degrees, which gives rise to electrical signals that are registered by the instrument. Various pattern recognition methods can then be employed to discriminate the unknown odors into predefined fingerprints. This potential to discern different odors at a significantly lower cost and time has fueled the research of utilization of e-noses in many fields. However, commercialized e-noses are still bulky, expensive, and offer a closed system that is very specific in terms of the sensor configuration and the aerosols that can be monitored, targeting niche markets with proprietary software. This thesis presents a novel portable, versatile, 3D Printed IoT connected e-nose, the advantages of which consist of its open source nature, ease of reproducibility and mass employability for a cost of around $200. The e-nose was tested in a wine classification exercise where different machine learning approaches were used, and the best one yielded a 99percent aroma prediction accuracy showing the potential of the proposed e-nose to service mass markets and compete to a large extent with commercial units.