[Kiaanaam]

Hi

I have these wavy solids and I want to simulate airflow passing through them. I have subtracted these solids from the fluid domain. The main geometry I have is around 40*40 cm and the bigger rectangle for fluid domain is only 50 cm longer at each side.

As the academic version does not allow for more than 512K mesh elements, how can I reduce this number? In my first try, I had around 46M elements !!! I tried reducing the number of divisions in the sizing. However, the element number increased to 49M.

How can I reduce the number to under 512K?
Should I suppress the wavy solids?

[vinerm]

Please post it in correct forum for faster and better response.

The approach is to use sweep or multizone mesh and not tetrahedral.

[Kiaanaam]

Hi Vinerm!
Thank you for your reply!

Should I use multizone mesh for the fluid domain (the big rectangle) or the wavy solids themselves?

[Kiaanaam]

Originally Posted by vinerm

Please post it in correct forum for faster and better response.

The approach is to use sweep or multizone mesh and not tetrahedral.

Hi Vinerm!
Thank you for your reply!

Should I use multizone mesh for the fluid domain (the big rectangle) or the wavy solids themselves?

[vinerm]

You need to keep the mesh of the solid region only if you want to model conjugate heat transfer. If not, then the solid region is not required. The mesh count is increasing because you have tetrahedral mesh in the fluid zone. Therefore, the suggestion was to use multizone in fluid zone. Solid zone should be hexahedral as much as possible.

[Kiaanaam]

Originally Posted by vinerm

You need to keep the mesh of the solid region only if you want to model conjugate heat transfer. If not, then the solid region is not required. The mesh count is increasing because you have tetrahedral mesh in the fluid zone. Therefore, the suggestion was to use multizone in fluid zone. Solid zone should be hexahedral as much as possible.

Hi Vinerm,
I tried the approach you suggested.
However, it shows the error: MultiZone blocking decomposition failed.
When I click on show problematic geometry, it shows the whole fluid domain that the geometry was subtracted from.

[vinerm]

As suggested earlier, I'd recommend you to post it in the relevant Forum.

[Kiaanaam]

Originally Posted by vinerm

As suggested earlier, I'd recommend you to post it in the relevant Forum.

I'm sorry, I thought this forum is for CFD and Fluent, where should I post my question exactly?

[RaiderDoctor]

Respectfully, Vinerm, I disagree with your suggestions. Depending upon what Kiaanaam is doing (strictly CFD or one-way FSI), using a tetrahedral mesh could be appropriate. Furthermore, it sounds as though Kiaanaam is a beginner and telling them to use multizone and sweep meshing strategies might only confuse them and not really solve the issue at hand. For tutorials on both, I'd recommend CFD Ninja on youtube (https://www.youtube.com/channel/UC3X...YZoh7LXY45YI6g). While not the most clear, it will give you decent exposure to these methods.


Now, on to your problem Kiaanaam. If I understand correctly, you are attempting to simulate heat transfer between fluid flow and six wavy structures. First question; why do you need to simulate all six? It looks as though this geometry is repetitive, and you can therefore take advantage of symmetric or periodic boundary conditions to simulate the other structures. This sort of thing is frequently done in CFD to help reduce computational cost while still maintaining accuracy. Remember that CFD is about solving a problem, and the key first step to solving any problem is to simplify the crap out of it. Just check your results thoroughly so that they match up with what you expect. If you need more info on this, I'll be happy to help.



Next question; why is your mesh crappy at the inlet and outlet, but super fine near the structures? This is not good meshing, as you may produce inaccurate results (I say *may* here because you will need to run a mesh sensitivity analysis to prove this). Think of it this way; you calculate a truly awful solution at the inlet, due to your terrible mesh. When you move to your structures, although your mesh is better (maybe), your results would still show a good approximation of a bad solution. Instead, try this strategy; begin by first meshing your entire fluid domain with uniform tetrahedral cells and show that your solution at the inlet and outlet are fairly decent. Then, iteratively increase the resolution near the structures until your results near exact from one refinement to the next. In practice, this is called a mesh sensitivity analysis, and should be the first thing done. Many people falsely assume that CFD is done in one go, which couldn't be further from the truth! CFD setup is a progressive process that is repeatedly built to perfection. Don't be afraid of that.

[Kiaanaam]

Originally Posted by RaiderDoctor

Respectfully, Vinerm, I disagree with your suggestions. Depending upon what Kiaanaam is doing (strictly CFD or one-way FSI), using a tetrahedral mesh could be appropriate. Furthermore, it sounds as though Kiaanaam is a beginner and telling them to use multizone and sweep meshing strategies might only confuse them and not really solve the issue at hand. For tutorials on both, I'd recommend CFD Ninja on youtube (https://www.youtube.com/channel/UC3X...YZoh7LXY45YI6g). While not the most clear, it will give you decent exposure to these methods.


Now, on to your problem Kiaanaam. If I understand correctly, you are attempting to simulate heat transfer between fluid flow and six wavy structures. First question; why do you need to simulate all six? It looks as though this geometry is repetitive, and you can therefore take advantage of symmetric or periodic boundary conditions to simulate the other structures. This sort of thing is frequently done in CFD to help reduce computational cost while still maintaining accuracy. Remember that CFD is about solving a problem, and the key first step to solving any problem is to simplify the crap out of it. Just check your results thoroughly so that they match up with what you expect. If you need more info on this, I'll be happy to help.



Next question; why is your mesh crappy at the inlet and outlet, but super fine near the structures? This is not good meshing, as you may produce inaccurate results (I say *may* here because you will need to run a mesh sensitivity analysis to prove this). Think of it this way; you calculate a truly awful solution at the inlet, due to your terrible mesh. When you move to your structures, although your mesh is better (maybe), your results would still show a good approximation of a bad solution. Instead, try this strategy; begin by first meshing your entire fluid domain with uniform tetrahedral cells and show that your solution at the inlet and outlet are fairly decent. Then, iteratively increase the resolution near the structures until your results near exact from one refinement to the next. In practice, this is called a mesh sensitivity analysis, and should be the first thing done. Many people falsely assume that CFD is done in one go, which couldn't be further from the truth! CFD setup is a progressive process that is repeatedly built to perfection. Don't be afraid of that.

Thank you very much RaiderDoctor for the things you explained!!
Actually I just started using Fluent a few days ago and I've been following tutorials online to be able to do my simulation. So I know my questions are super easy and even stupid.

For your first question, I want to simulate the airflow passing through these wavy surfaces (heat transfer is also intended but at this point, I'm ok to just focus on airflow). You are right, I didn't know about symmetric settings. So should I reduce the wavy surfaces to only two instead of six?

For your next question, I honestly have no idea why it is finer near the wavy surfaces cause I didn't change any settings on that. I was expecting it to be uniform all over. But now that you said it, I will look into it and try to understand why this happened.

You suggested to "begin by first meshing your entire fluid domain with uniform tetrahedral cells". Should I suppress the wavy surfaces for this? Or even delete them? Cause the fluid domain and the geometry are already subtracted from each other.

This is my first time using a CFD software ever and that's why I have zero knowledge on even common information. Thank you for your help!

[RaiderDoctor]

Unfortunately, you aren't alone in your confusion. CFD, and FLUENT especially, are not for beginners. The learning curve is huge, and you really need to spend a few months getting acquainted with the theory and program before you try anything on your own. This is not necessarily your fault, as most advisors/professors/bosses don't quite grasp this. In essence, FLUENT was made by programmers, for programmers, but is used by engineers. So there's this massive disconnect. For more tutorials, check out (https://confluence.cornell.edu/displ...arning+Modules).


Yes, reduce to two instead of six. This will substantially reduce your fluid domain.


The mesh is finer near the wavy surface because you added in number of divisions to the surface boundaries, and set the number to 40. While this increases the spatial resolution (i.e., the number of cells across the surface) it does nothing for the gradient resolution (i.e., number of cells close to the surface).



Keep the wavy structures in there and mesh uniformly. Then refine the mesh around the wavy structures.

[Kiaanaam]

Originally Posted by RaiderDoctor

Unfortunately, you aren't alone in your confusion. CFD, and FLUENT especially, are not for beginners. The learning curve is huge, and you really need to spend a few months getting acquainted with the theory and program before you try anything on your own. This is not necessarily your fault, as most advisors/professors/bosses don't quite grasp this. In essence, FLUENT was made by programmers, for programmers, but is used by engineers. So there's this massive disconnect. For more tutorials, check out (https://confluence.cornell.edu/displ...arning+Modules).


Yes, reduce to two instead of six. This will substantially reduce your fluid domain.


The mesh is finer near the wavy surface because you added in number of divisions to the surface boundaries, and set the number to 40. While this increases the spatial resolution (i.e., the number of cells across the surface) it does nothing for the gradient resolution (i.e., number of cells close to the surface).



Keep the wavy structures in there and mesh uniformly. Then refine the mesh around the wavy structures.

Thank you for your explanation!

Actually I'm a building technology student (with architecture background). That's why this type of simulation is not commonly used for us.

But my thesis is about airflow passing through complex geometries and that's why I'm using Fluent or at least I'm trying to....

I really hope I can find the right settings soon cause I don't have a few months to just learn the software

[Kiaanaam]

Originally Posted by RaiderDoctor

Unfortunately, you aren't alone in your confusion. CFD, and FLUENT especially, are not for beginners. The learning curve is huge, and you really need to spend a few months getting acquainted with the theory and program before you try anything on your own. This is not necessarily your fault, as most advisors/professors/bosses don't quite grasp this. In essence, FLUENT was made by programmers, for programmers, but is used by engineers. So there's this massive disconnect. For more tutorials, check out (https://confluence.cornell.edu/displ...arning+Modules).


Yes, reduce to two instead of six. This will substantially reduce your fluid domain.


The mesh is finer near the wavy surface because you added in number of divisions to the surface boundaries, and set the number to 40. While this increases the spatial resolution (i.e., the number of cells across the surface) it does nothing for the gradient resolution (i.e., number of cells close to the surface).



Keep the wavy structures in there and mesh uniformly. Then refine the mesh around the wavy structures.

Regarding the last point you mentioned, if I understood you correctly, you meant to not put the number of divisions for the wavy edges?

Cause as you can see in the attachment here, I reduced the wavy structure to only two of them, to define the symmetry axis later, and I didn't define a number of division, but still it is much finer around them.

[Kiaanaam]

Originally Posted by RaiderDoctor

Unfortunately, you aren't alone in your confusion. CFD, and FLUENT especially, are not for beginners. The learning curve is huge, and you really need to spend a few months getting acquainted with the theory and program before you try anything on your own. This is not necessarily your fault, as most advisors/professors/bosses don't quite grasp this. In essence, FLUENT was made by programmers, for programmers, but is used by engineers. So there's this massive disconnect. For more tutorials, check out (https://confluence.cornell.edu/displ...arning+Modules).


Yes, reduce to two instead of six. This will substantially reduce your fluid domain.


The mesh is finer near the wavy surface because you added in number of divisions to the surface boundaries, and set the number to 40. While this increases the spatial resolution (i.e., the number of cells across the surface) it does nothing for the gradient resolution (i.e., number of cells close to the surface).



Keep the wavy structures in there and mesh uniformly. Then refine the mesh around the wavy structures.

I also noticed that two of them are not meshed with tetrahedrons. How is this possible when the surfaces are exactly the same and just the direction is different?

[RaiderDoctor]

Correct. If you look on the left hand side of the image you attached, you'll see that one of the settings that is set to "yes" is "Capture Curvature". This is a nifty little feature that will automatically detect curves in your geometry and add more cells to that location, as fluid flow here tends require higher resolution. I strongly recommend you check out the manual and learn what each of these settings are, and what they affect.

[Kiaanaam]

Originally Posted by RaiderDoctor

Correct. If you look on the left hand side of the image you attached, you'll see that one of the settings that is set to "yes" is "Capture Curvature". This is a nifty little feature that will automatically detect curves in your geometry and add more cells to that location, as fluid flow here tends require higher resolution. I strongly recommend you check out the manual and learn what each of these settings are, and what they affect.

Thank you! I will look into it

[AlexanderZ]

select Mesh in project tree -> select any surface -> right click -> insert sizing -> put element size in m2 (it will be min element size) -> select Behavior hard

when you fill size value, press enter. Circle attached to mouse arrow will show you size of element, which you've defined