Ships are larger and slower than other means of transport because they carry a large amount of cargo. They are also operated at sea, where immediate access to land is restricted during normal voyages. In addition, they sail in extreme marine environments under the influence of rapidly changing weather. Therefore, studies are required to introduce advanced maintenance technologies into a ship to enable efficient and economical maintenance while maintaining the ship in the best operating condition.
Hence, studies are being conducted to apply predictive maintenance technologies to ship propulsion and generator engines for improving ship maintenance efficiency and economic feasibility, which enables preemptive maintenance by monitoring the condition of machines and systems and predicting abnormal signs. However, the research has several limitations, and it is challenging to derive practical results because of the character of ships that undergo preventive maintenance, making it difficult to collect normal and abnormal data.
In this study, a learning database was constructed by selecting and analyzing data that could determine the abnormality symptom of ship engines to apply the predictive maintenance to ship engines. In addition, by analyzing machine learning algorithms to be used for failure diagnosis and abnormality symptom prediction, we developed and verified algorithms for predicting the abnormality symptom and maintenance time of propulsion and generator engines. Furthermore, a generator engine model was built and verified to secure the normal and abnormal operation data of the engine via engine simulation. Through this, a foundation system was established to apply predictive maintenance to ships.
The predictive maintenance algorithm developed in research consists of engine condition criteria factors (main engine condition criteria value : MCCV, generator engine condition criteria value : GCCV) defined to improve engine abnormality symptom detection performance and algorithm processing speed as well as to derive intuitive results to be used in real ships.