Locomotive fault prognosis via machine learning
Πρόγνωση βλαβών κινητήριων μονάδων τροχαίου υλικού με μηχανική μάθηση
Μεταπτυχιακή διπλωματική εργασία
AdvisorNikolaou , Grigoris
KeywordsBearings ; BiLSTM ; K-means ; Locomotives ; Modelling ; Prognostics ; Regression ; Vibrations ; Condition-based maintenance ; Deep learning ; Fault prognosis ; Genetic algorithms ; Linear regression ; Multi-class classification ; Predictive maintenance ; Rolling stock ; Signal processing ; Supervised machine learning ; Μηχανική μάθηση ; Γενετικοί αλγόριθμοι ; Διάγνωση σφαλμάτων ; Ρουλεμάν ; Πρόγνωση σφαλμάτων ; Προγνωστική συντήρηση
Railway has played a vital role in transportation of both goods and people historically, continuing to hold an important share of the market with great potentials. Locomotives are the main part of a train and as with any mechanism; maintenance and troubleshooting are of critical importance. Except for corrective and preventing maintenance, the new trend in all industries is the fault prognostics, also known as predictive maintenance, whose goal is to detect an upcoming breakdown. Almost every mechanical compartment uses some type of bearings. So, this research focus on two main pillars, the first part is about to review the available technical manual and to count the bearings used in locomotives, while the second part is about to construct a deep machine learning model for bearings fault diagnosis and prognosis based on secondary data.