Prediction of peak cylinder pressure of a four-stroke
marine diesel engine using neural networks

Γαλλιάκης, Ιάκωβος; Galliakis, Iakovos

dc.contributor.author	Γαλλιάκης, Ιάκωβος	el
dc.contributor.author	Galliakis, Iakovos	en
dc.date.accessioned	2022-12-12T10:06:26Z
dc.date.available	2022-12-12T10:06:26Z
dc.identifier.uri	https://dspace.lib.ntua.gr/xmlui/handle/123456789/56421
dc.identifier.uri	http://dx.doi.org/10.26240/heal.ntua.24119
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	Μηχανική Μάθηση	el
dc.subject	Νευρωνικά Δίκτυα	el
dc.subject	Πρόβλεψη Πίεσης	el
dc.subject	Μέγιστη Πίεση Κυλίνδρου	el
dc.subject	Machine Learning	en
dc.subject	Neural Networks	en
dc.subject	Diesel Engine	en
dc.subject	Κινητήρας Ντήζελ	el
dc.subject	Pressure Prediction	en
dc.subject	Peak Cylinder Pressure	el
dc.title	Prediction of peak cylinder pressure of a four-stroke marine diesel engine using neural networks	en
dc.title	Πρόβλεψη μέγιστης πίεσης κυλίνδρου ενός τετράχρονου ναυτικού κινητήρα Diesel με χρήση νευρωνικών δικτύων	el
heal.type	bachelorThesis
heal.classification	Machine Learning	en
heal.classification	Μηχανική Μάθηση	el
heal.language	en
heal.access	free
heal.recordProvider	ntua	el
heal.publicationDate	2022-03-11
heal.abstract	In recent years, the demand for more efficient operation of engines has lead to an increase in the need for inexpensive and reliable monitoring tools. One parameter that is of great importance to the work producing process of an internal combustion engine is the in-cylinder pressure. The most common method for measuring such a parameter is through a piezoelectric pressure sensor; this solution however is quite expensive and the installation impractical and time-consuming. Others, more complex indirect methods include prediction of the pressure waveforms via utilisation of the acoustic emissions of the engine, or through the momentary crankshaft speed. A different approach to this task is explored through this thesis; the utilization of artificial neural networks, a machine learning model, that by processing easy to acquire data, namely the engine Speed, Torque, Lambda and Specific Fuel Consumption (BSFC), aims to make accurate predictions of the peak cylinder pressure. By using datasets from two different engines, both however being of the four-stroke, diesel type, two model groups were created; each grouped housed a large amount of different neural network architectures, in order to deduce the best hyperparameters for this task. After training and testing, it was concluded that the models were successful in predicting the peak pressure, as accuracy of 99.32% and 97.04% was reached by Model Set A and Set B respectively; it was also discovered that using the BSFC parameter as input worsened the performance of the models, leaving the engine Speed-Torque-Lambda as the optimal input vector. All the calculations and model building utilized the Julia programming language, and specifically the Flux machine learning package.	en
heal.advisorName	Papalambrou, George	en
heal.committeeMemberName	Kaiktsis, Lambros	en
heal.committeeMemberName	Grigoropoulos, Gregory	en
heal.academicPublisher	Εθνικό Μετσόβιο Πολυτεχνείο. Σχολή Ναυπηγών Μηχανολόγων Μηχανικών. Τομέας Ναυτικής Μηχανολογίας	el
heal.academicPublisherID	ntua
heal.numberOfPages	115 σ.	el
heal.fullTextAvailability	false