[mtrmasuko]

Hello everyone,

I'm trying to simulate ABL in a wind tunnel. I have the experimental data and I'm trying to validate my CFD results but I cannot get the same velocity profile.

I'm using k-epsilon turbulence model with enhanced wall treatment. I created mesh for y+ 30. 1st order solutions converged fastly but 2nd order solution gives "turbulent viscosity limited to viscosity ratio 10+e5" warning. When I refined my mesh, I got the same error again but lately.

You can see my mesh from below link:

https://ibb.co/JtPtpnT
https://ibb.co/wM8mRqN

I'm not sure that I have the correct turbulence model or wall fucntion for my case. Thanks for any help...

[LuckyTran]

Turbulence model and wall function should be okay. Even if it's inaccurate, you should be able to get a converged solution without errors.


If 1st order converges nicely but not 2nd order, and if the problem does not go away with just iterating longer, then that's a hint it's a mesh quality issue.Blind refinement can make the quality worse.


Did you do any wall clustering / prism layers? I can't see any, it's hard to tell.

[mtrmasuko]

Originally Posted by LuckyTran

Turbulence model and wall function should be okay. Even if it's inaccurate, you should be able to get a converged solution without errors.


If 1st order converges nicely but not 2nd order, and if the problem does not go away with just iterating longer, then that's a hint it's a mesh quality issue.Blind refinement can make the quality worse.


Did you do any wall clustering / prism layers? I can't see any, it's hard to tell.

After some refinement, I get rid off that warning message but this time continuity still not converging. I stopped solution after a while because, I don't think it will converge. Here it is my residuals:

https://ibb.co/0hL7CMv

Also, I took a snapshoot of my prisms... I'm not sure if you're asking this:

https://ibb.co/rFzdZvf


Thanks for reply...

[bpulat]

Looking at your residuals I think you should do more iterations.

It seems a bit early for me to say it is not converging.

And maybe a quick tip that can help you later is that you can always display cells with high turbulent viscosity ratio with iso-value.

Determine where it needs a mesh refinement if the problem is not because of your inputs or geometry but the mesh itself.

[mtrmasuko]

Originally Posted by bpulat

Looking at your residuals I think you should do more iterations.

It seems a bit early for me to say it is not converging.

And maybe a quick tip that can help you later is that you can always display cells with high turbulent viscosity ratio with iso-value.

Determine where it needs a mesh refinement if the problem is not because of your inputs or geometry but the mesh itself.

As you said I used iso-value for turbulent viscisity ratio and made refinement for that zones. Again continuity residual didn't converge even 10^-3.

This is the residuals: https://ibb.co/PTz0y25

I also checked mass flow and drag force. They reach a constant value. I used velocity inlet and pressure outlet boundary conditions. I really cannot figure out what is wrong.

Any help is appreciated.

[bpulat]

Do you have double precision on?

And can you check your input values if there is a typo or not?

Can you rerun the case with k-epsilon realizable if the standard is chosen?

[mtrmasuko]

Originally Posted by bpulat

Do you have double precision on?

And can you check your input values if there is a typo or not?

Can you rerun the case with k-epsilon realizable if the standard is chosen?

Yes double precision is on. I checked my inputs also. I will try realizable but before that I want to ask one more thing...

In the CAD, cubes are very close to the wall and y+ cells are getting smaller at that region as following:

https://ibb.co/0XH9sqn

Do you think this is the reason of the problem?

[bpulat]

Well, to eliminate if this affects the simulation or not decrease the number of cells across gap or refine that area.
But in my opinion, do these steps one by one to find out the cause.
There are lots of factors which might lead to an issue depending on the application of use.

[mtrmasuko]

OK. I will try them one by one. Thanks

[LuckyTran]

It might already be converged...

Residuals are a poor measure of convergence, especially on a bad mesh.

Make solution monitors and look at x,y,z velocity at some points in the domain and see how they behave.

Also the continuity residual is scaled by the worst residual in the 1st 5 iterations. So it's not a good idea ever to use continuity residual threshold as a convergence parameter, because it's scale dependent and initial guess dependent.

[mtrmasuko]

Originally Posted by LuckyTran

It might already be converged...

Residuals are a poor measure of convergence, especially on a bad mesh.

Make solution monitors and look at x,y,z velocity at some points in the domain and see how they behave.

Also the continuity residual is scaled by the worst residual in the 1st 5 iterations. So it's not a good idea ever to use continuity residual threshold as a convergence parameter, because it's scale dependent and initial guess dependent.

When I compared the results with experiment and they are pretty much different. That's probably because of mesh but I'm still suffering to fix it.. Thanks.