[Feature]: How does the boundary system work?

[sambaPython24]

Contact Details

No response

Is your proposal related to a problem?

Starting from the obstacle example, I saw that these are the boundaries that appear in the List

• None
• Anti_Bounce_Back_Outlet
• Bounce_Back_Boundary
• EquilibriumBoundaryPU

---

Some more insights would be very enlightening:

• None boundary : Why does the first boundary in the Simulation Object has to be a None object?
• Superposition of boundaries: Can I combine different boundaries and add them to that list? Are there any rules expect for that they should be not overlapping?
• Is the Bounce-Back-Boundary the default boundary for solid objects and Anti-Bounce-Back-Outlet for the flow to just leave the simulation grid?
• Is Equilibrium Boundary used to simulate an external, incoming stream?
• Why are the boundaries at the same time members of the simulation object and the flow object? Which ones are used for simulation?

Describe the solution you'd like

I would be delighted if we could have the rules for creating boundaries conditions in the documentation.
That would make it even easier to understand your great work.

Describe alternatives you've considered

Unify all boundaries to a single boundary_mask with values 0,1,2,3

Additional context

I would be delighted to be able to recreate a single like this simple fluid simulator here.

[McBs]

Hi @sambaPython24

I'm taking your suggestions on board. A comprehensive overhaul of the documentation, especially concerning boundary conditions and their application rules, is already planned. However, due to current projects, I won't be able to tackle this for about three months. I will address this afterwards.

But to help you right away, here are the answers to your specific questions:

1. The None Boundary
The None entry at the beginning of the list is a current internal implementation detail. During the simulation's initialization, this list is traversed to create an internal mask with integer values. Each integer value is then assigned to a specific boundary condition so that the simulation knows which logic to apply at each grid point. The first entry was used for collision

However, this is not an ideal solution and will be propably removed in an upcoming update. So, the assignment will start directly with the first element in the list, making the None entry unnecessary.

2. Superposition of Boundaries
A direct superposition of boundary conditions on the same grid points is not intended and would lead to undefined behavior. The rule here is: A grid point should only be influenced by one boundary condition.

If you need overlapping effects at specific locations, the recommended and safest approach is to write your own custom, combined boundary condition class. In this class, you can precisely define the logic and the order of operations for these special points.

3. Standard Boundaries and Their Usage
Bounce-Back-Boundary: Yes, this is the standard boundary condition for solid, impermeable walls (no-slip condition). We are also developing more advanced fluid-structure boundaries (e.g., interpolated and half-way bounce-back), which we plan to merge into the main branch later this year. If this is urgent we can also maybe help you the upcoming weeks. However our student who tested this BC is not at the office at the moment.

Anti-Bounce-Back-Outlet: I haven't configured the obstacle flow. Regarding its use in the obstacle flow example, that specific configuration may not be finalized. If you run into any issues with this boundary condition, I would recommend using one of these two alternatives for the outlet:

• Use the EquilibriumBoundaryPU condition, setting a target pressure/density.
• Use a Zou-He boundary condition, which is a common and robust choice. It extents the EquilibirumBoundary by a bounce back scheme of the higher order moments. I believe an implementation for this is already somewhere on my workstation. If you are interested in this approach, let me know, and I can provide you with the specifics in the upcoming days.

EquilibriumBoundaryPU: This boundary condition imposes a known macroscopic state (velocity U and pressure/density P) by setting the populations to their corresponding equilibrium distribution functions. In my experience, it's excellent for creating stable inlets or outlets. However, its main drawback is that it's not fully coupled with the internal flow dynamics (the populations are "forced" into equilibrium), which can make it less physically realistic compared to other methods.

I hope these explanations help you for now! Your suggestion to clearly document the rules for boundary conditions is exactly the right way to go, and as I mentioned, I will implement it.

Out of curiosity, what are you trying to simulate?
Don't hesitate to ask any further questions. Good luck with your project!

[sambaPython24]

First of all thank you very much for this very detailed answer. I think, you could already take this answer as a template for the
documentation along with the many other questions that have been answered. Also the comments in the code would be useful for that.

I am trying to simulate the turbulences around an object. In long term, I would like to extend the implementation by recent publications and your work gives a good fundament for that. At the moment, I am on my way to recreate the behavior provided in the video above (by an Eulerian approach).

Of course, I - and probably many more users - would be interested to see what you created for the interpolated, the half-way bounce-back and also for the Zou-He boundaries.

Maybe there is a possibility to create an experimental branch in this repo. If you continue to comment the functions and classes, the community might be able to help you with this. I hope to feel soon comfortable enough with the theory and the theory to join as well.

Thank you for your support and good luck with the projects !