Προβλεπτική συντήρηση πλοίων με χρήση χρονοσειρών
View/ Open
Keywords
ARIMA ; Προβλεπτική συντήρηση ; Συντήρηση ; Σύστημα προγραμματισμένης συντήρησης ; Χρονοσειρά ; Ανάλυση χρονοσειράς ; Γραμμικά μοντέλα πρόβλεψηςAbstract
This thesis deals with the analysis of time series in shipping data. This data was collected from a scheduled maintenance system of a shipping company and concerns the number of unscheduled failures that can occur each month on a machine installed on all the company's ships. The machine’s model name is CE-DE, 12V50DF. First, it is explained what a Planned Maintenance System is and why it is mandatory for all shipping companies to use it. Then, the four types of maintenance are being presented, which are the Corrective Maintenance, Preventive Maintenance, Predictive Maintenance and Condition Based Maintenance. It is explained what a time series dataset is and its characteristics are analysed. The time series which is used in this paper is a linear time series, therefore the linear models which are applied for prediction in linear time series are presented.
The purpose of this thesis is to predict the number of failures that may occur in the upcoming months on this machine. For this purpose, time series analysis methods were used and three ARIMA models with different parameters were tested. After each application of one of the ARIMA models, a prediction of the number of failures was made. Then according to the two information criteria, AIC and BIC, and the prediction tables of each model, the ARIMA models are compared to show which of the three models that were used, performs more effectively on the time series data set. The paper concludes the report by presenting the optimal ARIMA model for the specific time series used.