Development of ship performance models based on artificial neural networks and operational data

Anastasiou, Thalassinos; Αναστασίου, Θαλασσινός

dc.contributor.author	Anastasiou, Thalassinos	en
dc.contributor.author	Αναστασίου, Θαλασσινός	el
dc.date.accessioned	2022-06-14T07:02:10Z
dc.date.available	2022-06-14T07:02:10Z
dc.identifier.uri	https://dspace.lib.ntua.gr/xmlui/handle/123456789/55265
dc.identifier.uri	http://dx.doi.org/10.26240/heal.ntua.22963
dc.rights	Default License
dc.subject	Data science	en
dc.subject	Artificial neural networks	en
dc.subject	Machine learning	en
dc.subject	Deep learning	en
dc.subject	Performance monitoring	en
dc.title	Development of ship performance models based on artificial neural networks and operational data	en
heal.type	bachelorThesis
heal.classification	Ship performance	en
heal.language	en
heal.access	free
heal.recordProvider	ntua	el
heal.publicationDate	2022-03
heal.abstract	In this diploma thesis, two distinct types of data-driven models were developed: an artificial neural network and a multiple linear regression model. Both models attempted to forecast the fuel oil consumption of a crude oil tanker based operational data collected by an onboard monitoring system over an 18-month period. To ensure the reliability of the dataset, pre-processing is performed, which includes the removal of outlier data points via the imposition of thresholds and statistical filtering. After implementing data preprocessing, emphasis is placed on developing the artificial neural network with state-of-the-art training and optimization techniques to achieve the lowest possible error with a high degree of generalization capability. Then, the multiple linear regression model is developed, and both models are evaluated by computing critical metrics on an unknown dataset and by demonstrating their ability to construct fuel oil consumption - speed curves under a variety of loading and weather conditions. In both circumstances, the artificial neural network outperforms the multiple linear regression model, which is due to the presence of non-linearities in the physics of the problem. Python programming language has been used to carry out all of the processes in this thesis.	en
heal.advisorName	Themelis, Nicolaos	en
heal.committeeMemberName	Themelis, Nicolaos	en
heal.committeeMemberName	Spyrou, Konstantinos	en
heal.committeeMemberName	Zaraphonitis, Georgios	en
heal.academicPublisher	Εθνικό Μετσόβιο Πολυτεχνείο. Σχολή Ναυπηγών Μηχανολόγων Μηχανικών. Τομέας Μελέτης Πλοίου και Θαλάσσιων Μεταφορών	el
heal.academicPublisherID	ntua
heal.numberOfPages	91 σ.	el
heal.fullTextAvailability	false