Μελέτη διακύμανσης στρεπτικής ροπής για διάγνωση βλαβών σε ναυτικούς κινητήρες diesel

Μόκας, Στέφανος; Mokas, Stefanos

dc.contributor.author	Μόκας, Στέφανος	el
dc.contributor.author	Mokas, Stefanos	el
dc.date.accessioned	2020-05-14T08:28:25Z
dc.date.available	2020-05-14T08:28:25Z
dc.identifier.uri	https://dspace.lib.ntua.gr/xmlui/handle/123456789/50567
dc.identifier.uri	http://dx.doi.org/10.26240/heal.ntua.18265
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	Fault diagnosis	en
dc.subject	Marine diesel engines	en
dc.subject	Torsional vibrations	en
dc.subject	Modelica	en
dc.subject	Machine learning	en
dc.subject	Πρόβλεψη βλαβών	el
dc.subject	Ναυτικοί κινητήτες diesel	el
dc.subject	Στρεπτικές ταλαντώσεις	el
dc.subject	Γλώσσα προγραμματισμού Modelica	el
dc.subject	Μηχανική εκμάθηση	el
dc.title	Μελέτη διακύμανσης στρεπτικής ροπής για διάγνωση βλαβών σε ναυτικούς κινητήρες diesel	el
dc.title	Fault diagnosis of two-stroke marine diesel engines through torsional vibrations	en
heal.type	bachelorThesis
heal.classification	ΝΑΥΤΙΚΗ ΜΗΧΑΝΟΛΟΓΙΑ	el
heal.classification	MARINE ENGINEERING	en
heal.language	en
heal.access	free
heal.recordProvider	ntua	el
heal.publicationDate	2019-10-15
heal.abstract	The aim of this thesis is to provide a tool for the predictive maintenance of two-stroke marine diesel engines through monitoring of torsional vibrations and performance data. Costly failures can be avoided and downtime can be reduced with predictive maintenance, which is crucial for the vessel operation, sustainability and pro tability. In order to achieve this goal, a reliable modelling of the shafting system is very important. Such a model can be used for generating a dataset of the torsional vibrations during normal or faulty operation of a marine engine. This dataset can later be used for machine learning and training of a classi er model that would be able to discern between two classes of data: intact and faulty condition. Neural Networks (NN), Support Vector Machines (SVM) and Decision Trees are the most common algorithms used for classi cation based on machine learning. Within this thesis the Modelica language is used as the tool for developing the shafting system model and generating the torsional vibrations dataset. Emphasis is given on the presentation of Modelica language as a powerful tool for simulation and its use for torsional vibration analysis. Furthermore, within this study a literature review on the subject of predictive maintenance is conducted demonstrating that even though in other industries signi cant steps have been made on developing a diagnostics system, the maritime industry falls behind on the preventive maintenance of the marine diesel engines and thus there is still ample room for research and developments. A case of a container ship vessel driven by a two-stroke low-speed Diesel engine is studied. In particular, a 10,000 TEU Container Vessel's propulsion system was modelled, based on the existing torsional vibration analysis which was available within the frame of this study and simulations were conducted. The simulation requires that the model represents accurately the dynamic behaviour of the system for correct transient torsional vibration calculations. The shafting system was modelled using the Modelica language. In addition to that, the torsional vibrations theory was utilized and the steps for developing the case study model are analysed.The natural frequencies and modes of the shafting system being studied are determined, and the forced torsional vibration response is then calculated. The forced torsional vibration stress curves are obtained from the calculated vibrations. The same work is carried out for the case of a cylinder mis re and the results are compared with the available ones for veri cation. Finally, data generated from simulations is used for training machine learning algorithms in order to classify between intact and faulty operation. The trained classi er is able to distinguish between the intact condition and the one of cylinder mis ring, based on the dataset features that were extracted from the torsional vibration signals of the developed model. Moreover, the classi er is able to locate the fault location, indicating the most probable cylinder and percentage of mis ring. The developed method shows promising results for further research on a predictive maintenance tool that can be used for marine diesel engines through monitoring of torsional vibrations. Furthermore, the combination of additional performance data could provide more accurate and precise predictions.	en
heal.advisorName	Παπαδόπουλος, Χρήστος	el
heal.committeeMemberName	Καϊκτσής, Λάμπρος	el
heal.committeeMemberName	Παπαλάμπρου, Γεώργιος	el
heal.academicPublisher	Εθνικό Μετσόβιο Πολυτεχνείο. Σχολή Ναυπηγών Μηχανολόγων Μηχανικών. Τομέας Ναυτικής Μηχανολογίας	el
heal.academicPublisherID	ntua
heal.numberOfPages	108 σ.	el
heal.fullTextAvailability	false