Free-slip and no-slip BC

[LasNikas]

I ran the dam_break_gate_2d.jl example with the BoundaryModelDummyParticles and a high resolution. I noticed that the floppy plate (again) bends too much:

The same holds for the BoundaryModelMonaghanKajtar with the artificial viscosity parameter alpha=0.01. But with alpha=0.02 the bending behavior is fine (more on this later).

I checked the process of the rising gate with different resolutions with the BoundaryModelDummyParticles.

The following image shows the ouput (red particles belong to the rising gate and blue ones to the fluid).

That made me sit up. I removed the viscosity term from the fluid-boundary interaction. We then have a free slip boundary and everything is fine as we can see in the following image (regardless of the resolution).

The findings are as follows:

• note that alpha needs to adjust for different resolutions to maintain same Reynolds numbers (since alpha=f(h) )
• Thus, the plate bends too much for higher resolutions, if the parameter alpha is not adjusted (tank walls act like free slip boundaries --> no friction for the fluid). (not proper verified yet)
• Artificial viscosity is for the fluid-fluid interaction to reduce spurious flow oscillations. However, we can also use it for BoundaryModelMonaghanKajtar.
• But we must not use artificial viscosity for the Dummy Particle Model. This is phyicaly incorrect.

I will create a PR #100 in which I implement the no-slip BC for the BoundaryModelDummyParticles. It is based on the AdamiPressureExtrapolation from Adami 2012.

Here are the first outcomes for the no-slip BoundaryModelDummyParticles: the classical fluid at rest surrounded by two walls moving in negative x-direction.

[sloede]

Nice investigation and summary 👍 Also, the fluid at rest example looks a lot like it could be used for an open boundary experiment later (channel flow), doesn't it?

[LasNikas]

Correct! For the no-slip BC we need the velocity extrapolation as well. This PR is really crucial for the open boundary PR

[LasNikas]

A quick update: I plotted the x-deflection of the floppy plate for various boundary models (Dummy Particles and Monaghan model). Also, I varied the alpha (viscosity parameter for the artificial viscosity).
Note that Dummy Particle model is free slip here and the Monaghan model is ("artificial") no-slip. The green curve is from real experiment data. We can see that the no-slip condition slows down the fluid.

But!! That doesn't mean that the impact of the fluid is less as I thought yesterday. Though the fluid is slower, it rears up and the impact is even harder.

I applied the no-slip condition (blue particles) for the Dummy Particle model and compared it with the free-slip condition (red particles). As we can see in the images below.

Thus, the plate bends way too much for the no-slip condition.

[efaulhaber]

Closed by #169?

Can you please redo the experiments for verification?

[LasNikas]

closed by #169 and #170