The TwinShip consortium has published its latest Data Management Plan, detailing the datasets, data architecture and security measures supporting the project’s AI and Digital Twin applications for ship operations.
The document is the second version of its annually updated documentation. The first and second versions are both now available via ResearchGate
The project will combine operational data from Grimaldi RoRo vessels, Stena Line RoPax ferries, Stena chemical tankers, port operations at the Port of Valencia, weather data from Copernicus, ship design data from Knud E. Hansen, and green fuel and propulsion technology information from Wärtsilä and Bound4Blue.
Operational vessel data includes navigation, propulsion, fuel consumption, machinery and environmental measurements collected at intervals ranging from every two seconds for Grimaldi vessels to 14 seconds for Stena’s RoPax ferry and hourly for the Stena chemical tanker.
The datasets are intended to support vessel performance modelling, fuel consumption analysis, operational state identification and Digital Twin development.
The project will also generate annotated knowledge graphs that transform vessel datasets into RDF-based semantic models representing ship systems, machinery, propulsion equipment and their relationships. Rather than storing complete operational time series, the knowledge graphs will contain aggregated operational information to support AI workflows, decision-support applications and Digital Twin analytics.
A separate clustering dataset will classify vessel operating conditions using Gaussian Mixture Models and Expectation Maximisation algorithms. These operational processes will be used for applications including anomaly detection, missing-data recovery, adaptive state estimation and localised Digital Twin models.
The primary data platform will be the open-source VesselAI platform, which incorporates a federated data space for maritime data sharing. TwinShip will extend the platform with a Vessel Data Cloud Space to manage relationships between datasets using knowledge graphs and ontologies.
According to the plan, vessel datasets containing commercially sensitive operational information will remain restricted to authorised project partners, while selected non-confidential datasets will be published through repositories including DataverseNO, Zenodo, OpenAIRE and Open Research Europe in accordance with FAIR data principles.
