Physics Based, Data Augmented Digital Twins | Syroco

At the core of Syroco EfficientShip, the digital twin is a virtual representation of a real-world ship. The twin is used for simulation of efficiency and performance in a variety of configurations, based on design (propulsion, hull, etc.), operational (speed, trim, etc.) and environmental (wind, swell, etc.) parameters.

Each digital twin is built through the assembly of models, each of them representing the physics-driven behaviour of an element of the ship, including for example:

• Forces created by interaction of the hull with water, waves and swell
• Propulsion forces as well as drag, heeling and leeway caused by wind
• Efficiency of the engine (diesel, LNG, hydrogen...), propellers, energy consumption
• Drag created by the rudder, stabilisers and other appendices

The digital twin is an indistinguishable representation of the ship, and especially of her physics-driven behaviour, in specific conditions. In order to achieve this high degree of fidelity, each model used in the twin must accurately represent the behaviour of the associated component.

Each model has two tightly intertwined facets:

• On one side, the theoretical behaviour of the component, as computed based on design and the physics that apply (balance of forces, hydrodynamics, aerodynamics, etc.). This physics-driven behaviour can be obtained by a variety of techniques that include Computational Fluid Dynamics (CFD), Fluid Structure Interaction (FSI), parametric models, etc.
• On the other side, the actual behaviour of the component, as measured through data capture in operational or experimental conditions using IoT sensors, manual data logging, or other measurement techniques.

In a perfect world, these two facets would be a perfect mirror of one another. But the world is not perfect... There are actually many reasons why the theoretical and actual behaviours may diverge:

• A lack of precision in the design
• A lack of accuracy in the measurements
• Differences between actual construction and design
• Wear and tear of components which alter their characteristics and performance
• Actual operating conditions that differ from projections

In order to achieve the required fidelity, using only one facet (theoretical or actual) is not sufficient. Only the combination of both brings maximum accuracy and relevance. This is how models are built in Syroco EfficientShip, with physics-based models that are augmented by data.

It is however understood that not all required elements can always be available from the get-go for building each model. Syroco EfficientShip supports several modelling techniques, making it possible to create digital twins of ships that are already in operation as well as digital twins of ships that are currently in the design office and "exist" only in CAD software.

It is therefore possible to build models following different paths:

• Combine theoretical and actual behaviour elements from the get go, if they are indeed available.
• Create a model based solely on physics, using for example CFD simulation or parametric design elements. This model will be augmented with operational data when it becomes available, typically when the ship begins operations.
• Create a model based solely on operational/experimental data, typically obtained from historical operations, and enrich this model with physics where needed to improve accuracy.

Of course, since a digital twin is composed of several models, these models can be of different natures: for example one can build a physics-based model for hull, use combined physics and data-based models for the rudder and wind-assisted propulsion systems, and a purely operational data-based model for the engine and propeller.

At the end of the day, one should keep in mind that models (and therefore, twins) are never static and frozen. At any point during design and operations, data collected is fed into the models to recalibrate them and adjust their accuracy to actual operational performance.