Μηχανική μάθηση και Νανοτεχνολογία: Σύνδεση δομικών και λειτουργικών παραμέτρων νανοδομημένων επιφανειών με νανοτραχύτητα

Στέλλας, Αντώνιος-Χαράλαμπος; Stellas, Antonios-Charalampos

dc.contributor.author	Στέλλας, Αντώνιος-Χαράλαμπος	el
dc.contributor.author	Stellas, Antonios-Charalampos	en
dc.date.accessioned	2019-05-13T07:49:14Z
dc.date.available	2019-05-13T07:49:14Z
dc.date.issued	2019-05-13
dc.identifier.uri	https://dspace.lib.ntua.gr/xmlui/handle/123456789/48685
dc.identifier.uri	http://dx.doi.org/10.26240/heal.ntua.9048
dc.description	Εθνικό Μετσόβιο Πολυτεχνείο--Μεταπτυχιακή Εργασία. Διεπιστημονικό-Διατμηματικό Πρόγραμμα Μεταπτυχιακών Σπουδών (Δ.Π.Μ.Σ.) “Μικροσυστήματα και Νανοδιατάξεις”	el
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	Νανοτεχνολογία	el
dc.subject	Μηχανική μάθηση	el
dc.subject	Τραχύτητα	el
dc.subject	Νανοεπιφάνειες	el
dc.subject	Ενεργός επιφάνεια	el
dc.subject	Nanotechnology	en
dc.subject	Machine Learning	en
dc.subject	roughness	en
dc.subject	Active Area	en
dc.subject	nanosurfaces	en
dc.title	Μηχανική μάθηση και Νανοτεχνολογία: Σύνδεση δομικών και λειτουργικών παραμέτρων νανοδομημένων επιφανειών με νανοτραχύτητα	el
dc.title	Machine learning and Nanotechnology: Connecting the structural and functional parameters of rough nano-surfaces	en
heal.type	masterThesis
heal.classification	Νανοτεχνολογία	el
heal.classification	Μηχανική Μάθηση	el
heal.classification	Τεχνητή νοημοσύνη	el
heal.classification	Nanotechnology	en
heal.classification	Machine learning	el
heal.classification	Artificial intelligence	el
heal.classificationURI	http://skos.um.es/unescothes/C02659
heal.classificationURI	http://id.loc.gov/authorities/subjects/sh85079324
heal.classificationURI	http://skos.um.es/unescothes/C00261
heal.language	el
heal.access	free
heal.recordProvider	ntua	el
heal.publicationDate	2018-10-26
heal.abstract	In this diploma thesis, we investigate how artificial intelligence methods can assist the complex process of manufacturing, characterization and the use nanostructured materials. This research is an attempt to reinforce the two-way relationship between nanotechnology and artificial intelligence, where the latter will not only benefit from the first by using nanoelectronics but also enhance its further development by efficient analyzing its nano-data. Specifically, we use methods of machine learning (subfield of artificial intelligence) in order to connect the structural and functional parameters to nanostructured surfaces with roughness. Optimizing this link is crucial for further developments of nanotechnology and, above all, for the transformation of its results into widespread nanoproducts. The main motivation of this study is the stochastic character of surfaces with roughness that leads to a large number of parameters used to describe the roughness of a surface, however it is not clear in many cases their connection with a particular desirable functionality. The use of machine learning for the purpose of our work, for the purpose of diplomacy, would facilitate the prediction of functional parameters by the structural ones. At the same time, it could extract internal attributes from this link by asking questions such as: Can machine learning models demonstrate which structural parameters are of greater importance for this connection? In particular, our study focuses on the parameter of the active area, which affects the functionality in many applications (catalytic behavior, bio-structure, hydrophobicity, reflectivity, etc.), seeking its dependence on structural parameters of height and width roughness (Rms, correlation length, skewness, kurtosis). In the first part of our work we study the dependence of the active area on the structural parameters by means of simulated surfaces with roughness. In the second part we check whether machine learning can be used to make this connection by using model selective simulation results of the first part. From the results of this work we can conclude that models of machine learning based on the method of random forests and deep neural networks with differences between them manage to predict the precision of the active surface through the structural parameters. In addition, in the method of deep neural networks, this precision is maintained on a small number of training surfaces, making the method the most suitable for realistic applications. Finally, almost all methods show us that the Rms is the most critical parameter for predicting the active surface with a second correlation length, a finding that is confirmed and agrees with the simulation methods of the first part of the study.	en
heal.advisorName	Κωνσταντούδης, Βασίλης	el
heal.committeeMemberName	Τσουκαλάς, Δημήτριος	el
heal.committeeMemberName	Παπαδόπουλος, Γεωργιος	el
heal.committeeMemberName	Κωνσταντούδης, Βασίλης	el
heal.academicPublisher	Εθνικό Μετσόβιο Πολυτεχνείο. Σχολή Εφαρμοσμένων Μαθηματικών και Φυσικών Επιστημών	el
heal.academicPublisherID	ntua
heal.numberOfPages	105 σ.	el
heal.fullTextAvailability	true