Smart Gas Sensors: Deep Learning for the identification and classification of various gaseous species by sensors

Ανδρίκος, Κωνσταντίνος; Andrikos, Konstantinos

dc.contributor.author	Ανδρίκος, Κωνσταντίνος	el
dc.contributor.author	Andrikos, Konstantinos	en
dc.date.accessioned	2021-09-08T05:43:27Z
dc.date.available	2021-09-08T05:43:27Z
dc.identifier.uri	https://dspace.lib.ntua.gr/xmlui/handle/123456789/53813
dc.identifier.uri	http://dx.doi.org/10.26240/heal.ntua.21511
dc.description	Εθνικό Μετσόβιο Πολυτεχνείο--Μεταπτυχιακή Εργασία. Διεπιστημονικό-Διατμηματικό Πρόγραμμα Μεταπτυχιακών Σπουδών (Δ.Π.Μ.Σ.) “Υπολογιστική Μηχανική”
dc.rights	Αναφορά Δημιουργού-Μη Εμπορική Χρήση-Όχι Παράγωγα Έργα 3.0 Ελλάδα	*
dc.rights.uri	http://creativecommons.org/licenses/by-nc-nd/3.0/gr/	*
dc.subject	Smart gas sensing	en
dc.subject	Metal oxide gas sensors	en
dc.subject	Pattern recognition	en
dc.subject	Machine learning	en
dc.subject	Neural networks	en
dc.title	Smart Gas Sensors: Deep Learning for the identification and classification of various gaseous species by sensors	en
heal.type	masterThesis
heal.generalDescription	Η εργασίας εκπονήθηκε εξ' ολοκλήρου στα Ινστιτούτα Πληροφορικής και Τηλεπικοινωνιών και Νανοεπιστήμης και Νανοτεχνολογίας του Εθνικού Κέντρου Έρευνας και Φυσικών Επιστημών (Ε.Κ.Ε.Φ.Ε.) "Δημόκριτος"	el
heal.classification	Gas Sensors, Machine Learning	en
heal.language	en
heal.access	free
heal.recordProvider	ntua	el
heal.publicationDate	2021-07-05
heal.abstract	The aim of this present thesis was the exploitation of Artificial Intelligence (AI) algorithms for the discrimination of various gaseous species based on their type and concentration. For the detection of the volatile organic compounds we used metal oxide semiconductor (MOX) gas sensors. A key feature of these sensors is the alteration of one or more of their physical properties upon exposure to a gas stimuli in a way that is possible to measure and quantify. In our experiments during the exposure of an array of sensors to a deoxidizing gas, changes in the resistance of each sensor were measured. The acquired data were multivariate time series since we measured the response of an array of three sensors. The gas sensor array (GSA) delivers a unique fingerprint upon exposure to a gas stimuli. The next stage consists of pre-processing the acquired time series in order to use a pattern recognition algorithm for the recognition of the acquired fingerprint. In this context we used machine learning (ML) and deep learning (DL) models which after sufficient training, they are capable of predicting the class of a future recording of the GSA. The experimental process includes an odour delivery system, consisting of a sealed chamber where the GSA was placed, a gas injection phase, and the measure of the responses of the GSA by the Keithley 2400 instrument. We constructed the GSA’s electrical circuit and also a circuit for the control of the acquisition by an Arduino microcontroller. This work demonstrates the utilization of ML and DL algorithms in the field of smart gas sensors. For this purpose we used the dataset created in the laboratory but also some relevant datasets freely available from the UCI Machine Learning Repository.	el
heal.advisorName	Λαγαρός, Νικόλαος	el
heal.committeeMemberName	Λαγαρός Νικόλαος	el
heal.committeeMemberName	Θεοδώρου Δώρος	el
heal.committeeMemberName	Ριζιώτης Βαίλης	el
heal.academicPublisher	Εθνικό Μετσόβιο Πολυτεχνείο. Σχολή Χημικών Μηχανικών	el
heal.academicPublisherID	ntua
heal.fullTextAvailability	false