How in-service sensor data is redefining ship performance modelling

This paper resulted from a collaboration between Toqua and Malte Mittendorf (Technical University of Denmark - DTU).

Sea trial curves & theoretical weather correction factors are a common tool used to assess the performance of ships, but they have practical limitations that can impact their usefulness in operational settings. In this paper, we examine these limitations and suggest improvements to overcome these limitations.

First, we explore the possible improvement of creating calm water curves using operational data. Secondly, we also benchmark various wave correction factors and demonstrate that they yield comparable results regardless of the complexity of the formula used. Furthermore, we investigate the potential of ML to outperform traditional weather correction factors in ship performance modeling. We argue that ML, and specifically physics-informed machine learning, is the logical next step in the evolution of ship performance modeling given the increasing availability of data, computing power, and ML expertise.
The paper can be downloaded using the button above or accessed via ResearchGate.

Note: The ML models used in this paper are not as advanced as Toqua's own Ship Kernels. They do however clearly show the potential of ML in combination with in-service sensor data to improve ship performance modeling.