Marine combustion engines play a critical role in ship operations, where reliability and availability are essential for safe and efficient transportation. Unexpected engine failures can result in operational disruptions, costly downtime, schedule recovery expenses, increased fuel consumption, and high maintenance costs. Although modern vessels continuously generate large amounts of operational and maintenance data, much of this information is still underutilized in daily maintenance planning and technical decision-making. At the same time, recent developments in digitalization, IoT technologies, cloud computing, artificial intelligence, and Digital Twin systems create new possibilities for improving condition monitoring and predictive maintenance within the maritime industry.
This PhD project investigates how data-driven maintenance strategies can support predictive maintenance and condition monitoring of marine combustion engines. The research focuses on understanding how operational and maintenance data from propulsion systems can be collected, structured, and analyzed to support improved maintenance planning and earlier detection of engine degradation. An important part of the project is the investigation of the technical, organizational, and practical factors that influence the implementation of predictive maintenance solutions in real maritime environments. The project also investigates how operational datasets, maintenance records, and sensor data can be integrated into scalable analytical frameworks capable of supporting traceability and long-term maintenance analysis.
The research further investigates how AI and deep-learning methods can be applied to monitor engine performance and detect abnormal operating behavior in real time. Different anomaly detection approaches and residual analysis methods are explored to identify early signs of wear, degradation, and potential component failures before critical breakdowns occur. The project also investigates continual model retraining approaches that allow monitoring models to adapt as operational conditions, engine behavior, and available datasets evolve over time. Particular attention is given to the practical use of Digital Twin and AI-based solutions for supporting technical management and maintenance decision-making in maritime operations.
In addition, the project investigates how maintenance performance indicators and operational KPIs can be used to evaluate maintenance effectiveness, reliability, availability, downtime, and operational impact. The objective is to better understand how predictive maintenance strategies can contribute to reducing unexpected failures, improving engine uptime, supporting maintenance optimization, and reducing operational and maintenance-related costs. The research is carried out in collaboration with ShippingLab, Everllence, and Navigator Gas, where operational vessel and engine datasets are used to support model development, validation, and industrial implementation studies.