This paper was originally presented in March 2024 at the HullPIC Conference titled 'How to consistently select the right ship performance model in a fleet with mixed data availability'.
Abstract
Modeling ship performance can be done in many different ways. The spectrum includes purely theoretical formulas, purely data-driven models, and everything in between. With different data available for different vessels, how does one make the right choice for a whole fleet? This paper proposes a framework to select the best model in a consistent way, over a whole fleet where certain vessels may or may not have sea trial data, model tests, noon reports, sensor data, etc.
Conclusion
This paper explores the potential of orchestration to tackle the heterogeneity present across performance data and modeling approaches within the domain of ship performance. To capture the full potential of the data across a fleet, one must look beyond a single modeling approach and data type, and develop a holistic fleet-wide approach that’s able to address and overcome the different sources of heterogeneity. Otherwise, the potential of sensor-derived big data and machine learning to decarbonize the shipping industry will remain untapped.