Using machine learning algorithms to detect plasma disruptions in fusion reactors

Μαμαλουκάς, Πάτροκλος Μιλτιάδης; Mamaloukas, Patroklos Miltiadis

dc.contributor.author	Μαμαλουκάς, Πάτροκλος Μιλτιάδης	el
dc.contributor.author	Mamaloukas, Patroklos Miltiadis	en
dc.date.accessioned	2023-01-09T08:19:33Z
dc.date.available	2023-01-09T08:19:33Z
dc.identifier.uri	https://dspace.lib.ntua.gr/xmlui/handle/123456789/56545
dc.identifier.uri	http://dx.doi.org/10.26240/heal.ntua.24243
dc.description	Εθνικό Μετσόβιο Πολυτεχνείο--Μεταπτυχιακή Εργασία. Διεπιστημονικό-Διατμηματικό Πρόγραμμα Μεταπτυχιακών Σπουδών (Δ.Π.Μ.Σ.) “Φυσική και Τεχνολογικές Εφαρμογές”	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	Machine Learning Disruption	en
dc.subject	Plasma	en
dc.subject	Tokamak	en
dc.title	Using machine learning algorithms to detect plasma disruptions in fusion reactors	en
heal.type	masterThesis
heal.secondaryTitle	Χρήση Αλγορίθμων Μηχανικής Μάθησης για την Ανίχνευση Διασπάσεων Πλάσματος σε Αντιδραστήρες Σύντηξης	el
heal.classification	Plasma Physics, Artificial Intelligence	el
heal.language	en
heal.access	free
heal.recordProvider	ntua	el
heal.publicationDate	2022-10-01
heal.abstract	The aim of this thesis is to investigate how Machine Learning algorithms can be used to predict disruptions in fusion reactors. These events are called such because they disrupt a tokamak’s ability to confine plasma and cause a great deal of damage to the structure and equipment. Tokamaks are the leading design principle in fusion reactor design, so is ITER, the largest soon-to-be-built reactor ever conceived; thus, phenomena such as these must be prevented to the best of our abilities, to avoid compromising the future of energy production. Using bolometer data from the JET Tokamak reactor, we feed a 3-tier (Convolutional, LSTM and Linear Layers) Machine Learning model and iterate over sets of hyperparameters. Comparing different combinations of hyperparameters gives a qualitative perspective on the optimal configuration for our model and results are graded according to their f1-scores. Future improvements and optimizations are also suggested.	en
heal.advisorName	Κομίνης, Ιωάννης	el
heal.committeeMemberName	Κόκκορης, Μιχαήλ	el
heal.committeeMemberName	Αναγνωστόπουλος, Κωνσταντίνος	el
heal.academicPublisher	Εθνικό Μετσόβιο Πολυτεχνείο. Σχολή Εφαρμοσμένων Μαθηματικών και Φυσικών Επιστημών. Τομέας Φυσικής	el
heal.academicPublisherID	ntua
heal.numberOfPages	96 σ.	el
heal.fullTextAvailability	false