Fundamental Evaluation of Machine Learning-Based Digital Twin Systems for Predictive Maintenance of Large Engines in Maritime Settings

Machine learning-based digital twins or systems thereof have great potential to adequately model the behavior of complex systems and also directly provide valuable feedback to the physical entities. Digital twins of large internal combustion engines (ICEs) may, on the one hand, allow for a reliable depiction of the entire engine or key components. In combination with sensor readings, this depiction could be used for condition comparison and diagnosis. On the other hand, direct feedback to the engine or a component would allow for compensation measures, such as switching to a different control mode, if required, e.g., due to wear phenomena.

Wear and early failure indicators of engine components may also be used for preventive, condition-based maintenance or predictive maintenance (PdM) approaches. The latter is essentially based on predictions of the remaining useful life of an engine component based on its wear history. However, maintenance is not always feasible. For large ICEs used in maritime applications, constraints such as planned vessel routes and port times pose challenges for optimal predictive maintenance scheduling.

Target:

The goal of this Master’s thesis is to establish and evaluate a framework that determines the requirements for enabling PdM for large engines in a maritime setting using a machine learning–based digital twin. In particular, it shall be investigated how restrictions in maritime applications, such as planned routes and port times, can be taken into account in the context of optimal predictive maintenance scheduling. The appropriate use cases, constraints, and determining factors will be defined based on a thorough review of the literature and input from experts.

Tasks:

Literature research on digital twins for predictive maintenance applications (with particular focus on maritime ICE setting)
Definition of relevant use case(s) and scenarios
Determination and implementation of PdM approaches
Data generation (execution/simulation of scenarios defined)
Analysis of results and assessment of transferability to real application
Composition of the Master’s thesis

Your profile:

Interest in digital twins, machine learning, and optimization
Programming skills in Python or R
Experience in data analysis and machine learning
Experience in optimization is a plus
Field of study: Mechanical Engineering, Mechanical Engineering and Business Economics, Digital Engineering, Data Science, Computer Science, Information and Computer Engineering

Starting date: To be agreed

Duration: Approximately 6 months

Contact:

Dipl.-Ing. Christian Laubichler, +43 (316) 873-30089, christian.laubichler@lec.tugraz.at