Efficient dynamic ship-kite modeling

[1-Bart-1]

Hi everyone,

I’m looking for some input on building a dynamic ship model to support the optimization of kite flight paths and the realistic testing of flight controllers. My main requirements and open questions are below. Any advice, references, or practical experience would be greatly appreciated!

Goals:

• The kite dynamics should be coupled to the ship dynamics.
• The model should be fast enough for repeated optimization runs and controller-in-the-loop testing.
• It needs to approximate both planing and displacement regimes (ideally with some kind of interpolation between the two).
• Wave response is important: the ship should react to different wave conditions, at least at a basic level.

I would expect this model to be some sort of combination of:

• Rigid body dynamics for the ship movement.
• Hydrodynamics (which could be calculated using the Vortex Step Method).
• Wave modeling.
• Kite tether forces acting on the ship.
• Wind acting on the ship.
• Probably some things I missed.

Some questions:

• What level of model complexity is best for this use case? This is about balancing efficiency, while still getting the most important dynamics right.
• How do you model the planing behavior?
• How do you handle transitions between displacement and planing in simplified models?
• What’s the minimal set of environmental and control inputs needed for realistic behavior (waves, wind, rudder, etc.)?

[baptistelabat]

Hi Bart,

There are many possible options in model complexity. I would look at so called manoeuvrability models. I found this synthesis https://arxiv.org/pdf/2407.19673. I have already worked with Nomoto, Fossen and IEC 62065 models/simulators

Some of the models assume the following, which might be a good start.

• Neglect waves in front of propulsion and manoeuvring forces (both added resistance and dynamic)
• Neglect aerodynamic part in front of hydrodynamic part

However, future integration and adding kite force would require to use force based model which would rule out Nomoto's KT response model type or "black" box models. It might be possible to reuse some dimensionless coefficients model by using them to compute dimensionful forces.

Here would be my proposal:

• fully linearized models might not be enough in the frame of optimization, but quadratic should be enough.
• 3 degrees of freedom (2D in sea surface plane) might be enough as a start.
• assembly of appendages (rudder/daggerboard/foil) forces with analytical model and neglecting interaction as a start (why not Vortex step method if already used for the kite).
• consider only displacement or planing model (Savitzky seems yet to be the reference without CFD). Or use a look up table (anyway neglecting kite effect on lift/pitch)