Questions regarding Eddy3D simulation setup

[leeping120]

I have been learning and working with Eddy3D recently, and I have a few technical questions that I would greatly appreciate your guidance on:

1. Multi-direction simulation
   I noticed that a cylindrical domain is used for multi-direction simulations.
   Could you kindly explain how the inflow boundary conditions are defined in this case?
2. Turbulence settings
   Is it possible to adjust the k-ε values or turbulence length scale in Eddy3D?
3. Comparison with Butterfly
   I have read some comparisons between Eddy3D and Butterfly. Overall, Eddy3D seems significantly faster in simulation time. Is this mainly due to differences in the numerical methods, or are there other contributing factors?
4. Cylindrical domain and mesh refinement
   Regarding the cylindrical domain:
   a. Is the mesh in this domain still orthogonal, or does it take another shape (such as tetrahedral/prismatic)?
   b. Compared with a rectangular domain, is there a fundamental difference in the way the mesh is generated and the solver
   processes it, since in rectangular grids the computation direction is more visually evident?
5. Mesh refinement details
   For mesh refinement, I understand that Eddy3D can refine around the building geometry.
   a. Is it possible to specify a refinement region (e.g., within 6 meters of the building)?
   b. In areas farther from the building, is it only possible to keep the same mesh size, or can a gradual coarsening
   (progressive mesh grading) also be applied?
   Thank you very much for your time and for developing such a powerful tool. Your insights would be extremely valuable to my research.

[kastnerp]

I have been learning and working with Eddy3D recently, and I have a few technical questions that I would greatly appreciate your guidance on:

1. Multi-direction simulation
   I noticed that a cylindrical domain is used for multi-direction simulations.
   Could you kindly explain how the inflow boundary conditions are defined in this case?

Please check here and here:
https://www.researchgate.net/publication/338030136_A_Cylindrical_Meshing_Methodology_for_Annual_Urban_Computational_Fluid_Dynamics_Simulations
https://www.researchgate.net/publication/356780551_Eddy3D_A_toolkit_for_decoupled_outdoor_thermal_comfort_simulations_in_urban_areas

2. Turbulence settings
   Is it possible to adjust the k-ε values or turbulence length scale in Eddy3D?

Not in Eddy3D, but you can in OpenFOAM. We are working on a rewrite from scratch, the next major version of Eddy3D will allow you to overwrite any arbitrary values.

3. Comparison with Butterfly
   I have read some comparisons between Eddy3D and Butterfly. Overall, Eddy3D seems significantly faster in simulation time. Is this mainly due to differences in the numerical methods, or are there other contributing factors?

Likely due to differences in the numerics and the meshing. The solvers should be identical.

4. Cylindrical domain and mesh refinement
   Regarding the cylindrical domain:
   a. Is the mesh in this domain still orthogonal, or does it take another shape (such as tetrahedral/prismatic)?
   b. Compared with a rectangular domain, is there a fundamental difference in the way the mesh is generated and the solver
   processes it, since in rectangular grids the computation direction is more visually evident?

Please see here: https://www.researchgate.net/publication/338030136_A_Cylindrical_Meshing_Methodology_for_Annual_Urban_Computational_Fluid_Dynamics_Simulations

5. Mesh refinement details
   For mesh refinement, I understand that Eddy3D can refine around the building geometry.
   a. Is it possible to specify a refinement region (e.g., within 6 meters of the building)?

AccBBox does that, but it does not specify 6m exactly. Again, you will be able to do this in the next major version.

b. In areas farther from the building, is it only possible to keep the same mesh size, or can a gradual coarsening
(progressive mesh grading) also be applied?

You could do this with multigrading, but currently not supported by Eddy3D. https://www.youtube.com/watch?v=dmHvNKqZ5qU

Thank you very much for your time and for developing such a powerful tool. Your insights would be extremely valuable to my research.

Thanks, we try our best to make it better over time :). Unfortunately, we are constrained in time and funding, so progress is slow.

[leeping120]

Could you please explain why the cylindrical simulation domain converged faster than the box-shaped domain in your study?🙏

[kastnerp]

If I remember correctly it had fewer cells, given the best practice guidelines. Convergence is a function of many variables (cell size, cell quality, boundary conditions, and many more). I don't think one can point to a single reason why it converges faster.

@karadagi recently fixed some issues we had with this domain in version 0.4.15.10. I would appreciate it if you could let us know if it works well for you

[leeping120]

I was wondering if the validation of the simulations has also tested the accuracy of the pressure coefficient (Cp) results, or if it only focused on wind velocity.
In my case, the Cp results seem quite inaccurate, especially on models with concave or convex surfaces. Have you encountered similar issues before?🙏

[kastnerp]

We compared a standard tower geometry and it looked fine. I have not looked at anything else. Feel free to post your results here (including the GH file), I hope we find time to take a look.

[leeping120]

1022test.zip
🙏🙏🙏

[karadagi]

Hi, the actual building geometry and the one created as a mesh on OpenFOAM side differs, so the probe locations are really important. You can try to offset them a bit to make sure the pressure values are not infinite. You can also try to visualise the results at ParaView side if you are familiar.