Bayesian Update on Nonlinear Multiscale Systems

Πυριαλάκος, Στέφανος-Χρήστος; Pyrialakos, Stefanos-Christos

dc.contributor.author	Πυριαλάκος, Στέφανος-Χρήστος	el
dc.contributor.author	Pyrialakos, Stefanos-Christos	en
dc.date.accessioned	2020-06-23T08:03:50Z
dc.date.available	2020-06-23T08:03:50Z
dc.identifier.uri	https://dspace.lib.ntua.gr/xmlui/handle/123456789/50811
dc.identifier.uri	http://dx.doi.org/10.26240/heal.ntua.18509
dc.description	Εθνικό Μετσόβιο Πολυτεχνείο--Μεταπτυχιακή Εργασία. Διεπιστημονικό-Διατμηματικό Πρόγραμμα Μεταπτυχιακών Σπουδών (Δ.Π.Μ.Σ.) “Υπολογιστική Μηχανική”	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	Bayesian Update	en
dc.subject	Nonlinear analysis	en
dc.subject	Finite element analysis	en
dc.subject	Multiscale systems	en
dc.subject	Μπειζιανή Αναβάθμιση	el
dc.subject	Μη γραμμική ανάλυση	el
dc.subject	Πεπερασμένα στοιχεία	el
dc.subject	Πολλαπλές Κλίμακες	el
dc.title	Bayesian Update on Nonlinear Multiscale Systems	en
heal.type	masterThesis
heal.classification	Υπολογιστική Μηχανική	el
heal.access	free
heal.recordProvider	ntua	el
heal.publicationDate	2020-03-04
heal.abstract	Mechanicals systems are often uncertain in their response under different conditions. Examples of such uncertainties include experimental errors, lack of data, uncertain model parameters, and systematic model inadequacy. This uncertainty is most of the times a hard task to define properly. The overall uncertainty can be decomposed on a number of different parameters that affect the system with several ways. These parameters are not always straight forward and the connection of their probabilistic structure with that of the whole system’s may not be possible to be found from beforehand. A characteristic example of this issue is the investigation of the uncertainty of the microstructure for heterogeneous materials. The direct investigation of the microstructure is a very challenging and costly task. This issue can be countered efficiently by taking advantage of the high amount of information and data that can be collected and stored while observing the system’s response macroscopically and with a multiscale approach update the random microscale parameters. The Bayesian framework, which combines new information with preexisting models, can be used in order to update the microscopic scales with acquired data from the investigation of either those scales or related macroscopic scales. Several studies have been done for the stochastic formulation of microstructures that ultimately define the macrostructure (i.e. random wrinkling of graphene sheets [16], stochastic geometric formulation of carbonate rocks [17]), without however any consideration of real data that can reformulate the microstructure. On the other hand, multiscale Bayesian updating has been implemented mainly on image processing studies (i.e. speckle removal in ultrasound images [18], image segmentation [19,20]). However, Bayesian updating on microscopic stochastic parameters of heterogeneous materials based on multiscale procedures have yet to be investigated thoroughly, in spite of the useful information that can be gathered from it. P a g e \| 2 Here the Bayesian updating of the probabilistic structure of the uncertain parameters describing the interaction between carbon nanotubes (CNTs) and the surrounding polymer matrix is considered. A MCMC technique with the use of the Metropolis Hastings is implemented on a semi-concurrent multiscale approach (FE2) in order to update the beliefs of the microscopic parameters according to a macroscale deformation observation. The posterior distributions of these parameters are then compared to deterministic solutions of the multiscale model in the direction of verifying the efficiency of the procedure.	en
heal.advisorName	Παπαδόπουλος, Βησσαρίων	el
heal.committeeMemberName	Παπαδόπουλος, Βησσαρίων	el
heal.committeeMemberName	Λαγαρός, Νικόλαος	en
heal.committeeMemberName	Σπηλιόπουλος, Κωνσταντίνος	el
heal.academicPublisher	Εθνικό Μετσόβιο Πολυτεχνείο. Σχολή Πολιτικών Μηχανικών	el
heal.academicPublisherID	ntua
heal.numberOfPages	48 σ.	el
heal.fullTextAvailability	false