IMPLEMENTING PREDICTIVE MAINTENANCE IN MARITIME OPERATIONS: A CASE STUDY WITH FRUGAL TECHNOLOGIES
Author
Michelena Banuelos, Endika
Term
4. semester
Publication year
2024
Submitted on
2024-06-02
Abstract
This thesis examines how meteorological conditions affect vessel performance in the context of implementing predictive maintenance (PdM). In a case study with Frugal Technologies, vessel operational data are combined with weather datasets to analyze fuel consumption and shaft power, which feed into the engine efficiency indicator, Specific Fuel Oil Consumption (SFOC). Following the CRISP-DM process, data are cleaned and enriched with new features, including relative wind and wave direction and SFOC. The study identifies distinct engine behaviors under varying weather and reconstructs a reference curve for ideal fuel use during near-calm water sailing. Building on this, a machine learning model (XGBRegressor) is trained and tested to predict extra fuel usage due to weather loads, enabling SFOC to be corrected for external factors and used more robustly in a PdM solution. The work aims to provide shipowners with timely insight into weather-driven overconsumption and represents a step toward more reliable condition monitoring, with potential applications in weather routing and voyage optimization. The research asks what influence meteorological conditions have on vessel performance and how weather impacts shaft power and fuel consumption.
Denne afhandling undersøger, hvordan meteorologiske forhold påvirker et fartøjs ydeevne i forbindelse med implementering af predictive maintenance (PdM). I et casestudie gennemført med Frugal Technologies kombineres skibsdriftsdata med vejrdatasæt for at analysere brændstofforbrug og akselkraft, som indgår i motorens effektivitetsindikator, Specific Fuel Oil Consumption (SFOC). Med udgangspunkt i CRISP-DM-processen renses og beriges data med nye features, herunder relativ vind- og bølgeretning samt SFOC. Studiet identificerer forskellige motoradfærd under skiftende vejr og genskaber en referencekurve for ideelt brændstofforbrug ved sejlads i næsten roligt vand. På denne basis trænes og testes en maskinlæringsmodel (XGBRegressor) til at forudsige ekstra brændstofforbrug forårsaget af vejrbelastninger, så SFOC kan korrigeres for eksterne faktorer og anvendes mere robust i en PdM-løsning. Arbejdet har til formål at give rederier løbende indsigt i vejrrelateret merforbrug og udgør et skridt mod mere pålidelig tilstandsovervågning med potentiale for anvendelser inden ruteplanlægning og rejseoptimering. Forskningsspørgsmålet er, hvilken indflydelse meteorologiske forhold har på fartøjets ydeevne, og hvordan vejr påvirker akselkraft og brændstofforbrug.
[This apstract has been generated with the help of AI directly from the project full text]