DEMOPS - Maskininlärningsbaserad modellering av hastighetseffekt för att minska bränslekostnader och utsläpp från frakt

Forskningsprojekt, 2020 – 2024

A ship's fuel consumption can increase significantly when sailing in harsh sea conditions. All measures to increase the ship's energy efficiency must rely on a detailed description of the ship's energy performance, ie. power-to-speed ratio, at sea.
Current theoretical physical models always contain large uncertainties in the description of a ship's energy performance, especially in the mechanical system models. Some blackbox performance models have been constructed using machine learning methods based on data on the ship's performance in different conditions. However, the Blackbox models are only useful for a specific vessel with data entered for the model design.
In this project we will develop functional data analysis algorithms (FDA algorithms) to select / simulate correct ship data. This data will be used for sophisticated machine learning algorithms to combine with theoretical models to better understand and construct models of ship energy performance. Some reverse machine learning algorithms will be developed to accurately describe the wave conditions encountered by a ship. Finally, these models will be demonstrated to show how they can be used to develop energy-efficient measures for ships. Potential fuel savings and reductions in air emissions will be identified through the demonstrations.