The primary objective of FUSE is to develop knowledge and methodologies for energy-efficient and safe ship operations by fusing AI and ship hydrodynamics that enable rapid emission reductions from the global fleet through next generation voyage optimization. The main idea is to develop theoretical approaches and methodology utilizing the potential of using AI in combination with classical hydrodynamics to develop more accurate numerical models for ship performance, including both energy and motion characteristics. Traditional hydrodynamic models are based on first-principle, reflecting physical laws. AI-based models which uses operational data are used to establish a statistical relationship between inputs and outputs. Both approaches have their pros and cons, and we aim to fuse the best of both approaches and develop hybrid models as input to voyage planning and onboard decision support systems. The project partners include SINTEF Ocean, SINTEF Nordvest, NTNU, Torghatten, Vard Electro, NAVTOR, and G2 Ocean.
During the first year of the project, we have collected user requirements for ship models and voyage optimization from the industry partners with a focus on two selected use cases; deepsea and ferry operations. In this regard, we have held a project meeting on one of the Torghatten ferries on the Halhjem-Sandvikvåg route to hear experiences and input from the captains. We have also started to collect and analyze operational ship data from the partners in the project. This applies to operational data from Torghatten ferries on the Halhjem-Sandvikvåg route south of Bergen and data from the deepsea ships G2 Ocean. In addition, we have received collected and processed data from Grieg Star's ships that are being worked on in another research project called Green AI for Sustainable Shipping. By receiving already collected and processed data from another project, we save a lot of time and work on this in FUSE, which allows us to focus more on developing methods and knowledge for hybrid (AI and ship hydrodynamics) models for ship performance. This August, a PhD candidate started at the Department of Marine Engineering at NTNU who will work on the project, which focuses on merging AI with ship hydrodynamics.
The FUSE vision is to rapidly reduce emissions from the global fleet by unlocking the full potential of operational data through fusing AI and ship hydrodynamics for next generation voyage optimization. The main idea is to develop theoretical approaches and methodology utilizing the potential of using AI in combination with classical hydrodynamics to develop more accurate numerical models for ship performance, including both energy and motion characteristics. Traditional hydrodynamic models are based on first-principle, reflecting physical laws. AI-based models which uses operational data are used to establish a statistical relationship between inputs and outputs. Both approaches have their pros and cons, and we aim to fuse the best of both approaches and develop hybrid models as input to voyage planning and onboard decision support systems. As a starting point, we will gather user requirements from the industrial partners focusing on two selected use cases; deepsea and ferry operations. The industry partners will be involved during the whole project period ensuring that the competence and results can be exploited into their product and services. The industry partners will contribute with operational data from their ships as research data.
The research activities will cover data analytics, AI and ship hydrodynamics and voyage optimization and decision support. Research partners are SINTEF Ocean, SINTEF Ålesund, and NTNU. A PhD candidate will be educated from the NTNU with supervisors from Marine Technology and Engineering Cybernetics with focus on fusing AI with ship hydrodynamics. Relevant research topics involve comparing the different approaches for estimating ship performance, considering accuracies and valid usage of the models and evaluation of sensitivity and applicability of the models. Finally, the potential impact of using hybrid models in voyage optimization will be investigated and demonstrated.
