Prediction of resistance of MARAD systematic series’ Hullforms using atificial neural networks

Margari, Vasiliki; Μαργάρη, Βασιλική

dc.contributor.author	Margari, Vasiliki	en
dc.contributor.author	Μαργάρη, Βασιλική	el
dc.date.accessioned	2018-01-10T12:26:45Z
dc.date.available	2018-01-10T12:26:45Z
dc.date.issued	2018-01-10
dc.identifier.uri	https://dspace.lib.ntua.gr/xmlui/handle/123456789/46189
dc.identifier.uri	http://dx.doi.org/10.26240/heal.ntua.15028
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	Ship resistance	en
dc.subject	Artificial neural networks	en
dc.subject	MARAD systematic series	en
dc.subject	Multi-layer perceptron	en
dc.subject	Radial basis function neural networks	en
dc.subject	Support vector machines	en
dc.subject	Αντίσταση πλοίου	el
dc.subject	Τεχνητά νευρωνικά δίκτυα	el
dc.subject	Συστηματική σειρά MARAD	el
dc.subject	Δίκτυα ακτινικών συναρτήσεων βάσης	el
dc.subject	Μηχανές διανυσμάτων υποστήριξης	el
dc.subject	Πολυστρωματικά δίκτυα	el
dc.title	Prediction of resistance of MARAD systematic series’ Hullforms using atificial neural networks	en
heal.type	bachelorThesis
heal.classification	Naval architecture	en
heal.language	en
heal.access	free
heal.recordProvider	ntua	el
heal.publicationDate	2017-10-25
heal.abstract	Maritime Administration (MarAd) in cooperation with Hydronautics Inc., motivated by the rising interest for full hull form merchant ships in the early 1970’s and the lack of systematic data regarding the performance of these vessels, developed the so-called MARAD Systematic Series, comprised of 16 full hullforms, specifically designed for use as bulk carriers and tankers. The experimental resistance data are available in a series of diagrams illustrating the residual resistance coefficient as a function of the geometric characteristics of each hull for a range of Froude numbers. The present thesis investigates the potential of Artificial Neural Networks (ANNs) to estimate the residual resistance coefficient, and subsequently the resistance, of hullforms designed according to MARAD Systematic Series, given their main dimensions and block coefficient. The data used for the training of the networks were collected from five diagrams, which are provided by the Series, and illustrate the resistance coefficient for different combinations of their length to breadth and breadth to draft ratios and block coefficient. To this end, three different types of ANNs have been considered; Multi-layer perceptron (MLP) networks, Radial Basis Function (RBF) networks, and Support Vector Machines (SVM). The performance of a network, regardless its type, is affected by its characteristics, among which its architecture and learning method are of particular importance. Several trials have been conducted for every type of network, changing systematically these characteristics. The developed networks have been thoroughly evaluated, assessing their potential to accurately predict the resistance of MARAD-type hullforms. A total number of 616 networks were developed; 380 MLPs, 125 RBFs and 111 SVMs. Selected alternatives of these networks, i.e. 4 MLPs, 2 RBFs and 2 SVMs are presented and their potential for the prediction of MARAD hullforms’ resistance is discussed. The deviations of the best-performing networks’ predictions from the resistance data provided by MARAD were not higher than 1.6% for hullform characteristics inside the limits of the training dataset and 6.8% outside them. ANNs are quick and effective in estimating the resistance of MARAD hullforms, eliminating the need for searching through the diagrams that provided their training data. The results indicate that their use could be successfully applied also in the case of other Systematic Series. In addition, it might be argued that ANNs could be successfully trained to estimate calm water resistance of selected hullform types, based on the results of systematic calculations using advanced CFD software tools.	en
heal.advisorName	Ζαραφωνίτης, Γεώργιος	el
heal.committeeMemberName	Ζαραφωνίτης, Γεώργιος	el
heal.committeeMemberName	Βεντίκος, Νικόλαος	el
heal.committeeMemberName	Σπύρου, Κωνσταντίνος	el
heal.academicPublisher	Εθνικό Μετσόβιο Πολυτεχνείο. Σχολή Ναυπηγών Μηχανολόγων Μηχανικών. Τομέας Μελέτης Πλοίου και Θαλάσσιων Μεταφορών	el
heal.academicPublisherID	ntua
heal.numberOfPages	81 σ.	el
heal.fullTextAvailability	true