Learning Forum

[Aras karimi]

Hi to all,

I am working on a 3D airfoil. Currently, I am doing a mesh independence study to assess the dependence of the simulation results.
For this purpose, I plot the two parameters CL and CD when calculating. My solver is Fluent and I use K-W SST turbulence model. My mesh is the structure that I produced in icem-cfd. 
I first started with a 552K cell mesh with  Y+ <1 , and the residuals converged very well. which is according to the following figure:

 

Now, I increased the number of grid cells to 1,263,000 cells, but unfortunately, this time, the residuals fluctuate strongly and do not intend to converge, and I also receive the message (Stabilizing pressure coupled to enhance linear solver robustness.) repeatedly when solving, which is as follows:

 

I really wonder why this is happening. To create a new mesh, I did not change the growth rate of the block lines at all, and only increased the number of nodes of the block lines, and at the end, I checked the mesh quality parameters, which were very good.

Mesh 1:  (has very good convergence)

Aspect Ratio: 1-5700

Determinant : 0.87-1

Orthogonal Quality: 0.82-1

Quality: 0.87-1

Skew: 0.69-1

 

                                            Mesh 2:  (has very bad convergence)

Aspect Ratio: 1-5660

Determinant : 0.895-1

Orthogonal Quality: 0.837-1

Quality: 0.87-1

Skew: 0.674-1

 

If you pay attention, you can see that even the quality of the second mesh is a little better than the first mesh, but its convergence is much worse and full of (Stabilizing pressure coupled to enhance linear solver robustness.) messages.

The aspect ratio of the network is as follows, which shows the elements behind the airfoil:

 

Now, I increased the mesh behind the airfoil and found that my non-convergence problem was solved:

 

See, with the increase in the number of cells behind the airfoil, my mesh has increased from 1,263,000 cells to 2,000,000 cells, which takes time to solve. If this amount of cell increase is applied on the surface of the airfoil, it gives much more accurate results than applying  at the back of the airfoil. 

In other words, I think that the mesh behind the airfoil does not need a significant increase to solve the convergence problem, and if this mesh increase is applied to the surface of the airfoil, I will get accurate results closer to the Experimental results.  Increasing the grid behind the airfoil solved my convergence problem though.

 

 

I have seen in many Paper that the mesh behind the airfoil is gradually enlarged to the end, which is as follows: But when I gradually enlarge the mesh behind the airfoil to the end, the flow does not converge:

 

 

I also used the following ratios on the lines of the blocks:

Are these Ratios suitable?

Why do I have to make the grid behind the airfoil so small to solve the problem of its non-convergence? Do you have a method or an idea to solve the problem of its high aspect ratio without increasing the cells behind the airfoil?

Thanks in advance

Regards

[Federico]

I would suggest locating the cells with worst residuals as your solution diverges by using cell registers to see where the issue is coming from. Compare this with your partitions to make sure they do not occur in regions with high gradients.

[Aras karimi]

 

How can I find cells with very low and bad residues? 

Look, dear Federico, it is very strange and unbelievable. I increased the first generated network (552,000 cells) only once and without changing the Ratio of the block lines until I reached the network of 1,263,000 cells. Now this network does not converge at all. Why is that ?   The quality of the network is also good and there are no problems.

 

[Aras karimi]

Dear Federico,

How can I find cells with very low and bad residues? 

Look, dear Federico, it is very strange and unbelievable. I increased the first generated network (552,000 cells) only once and without changing the Ratio of the block lines until I reached the network of 1,263,000 cells. Now this network does not converge at all. Why is that ?   The quality of the network is also good and there are no problems.

[Federico]

Hello Aras,

We have responded to all of the issues that you have highlighted and provided our recommendations to address them. Going beyond this would require troubleshooting of your specific case and is beyond the scope of this Forum. 

To view high residual cells, you can create cell registers > Field Variables > Residuals. 

[Rob]

Other comments here  /forum/forums/topic/mesh-independence-study-3/  Other thread now locked. 

[Aras karimi]

According to our discussion in the topic you locked, continue our conversation:

********* what you said :

The reduction in cell size may mean you're also picking up a flow feature you'd previously missed. It that's transient then, yes, it's potentially caused by refining the mesh. 

Cell count isn't always a good indicator for studies. The authors may have used different meshing strategies making direct comparisons difficult. 

 

If the residuals fluctuate by increasing the number of grid cells, and the solution does not converge, it is not a good thing. For me, after every repetition, the message "Pressur Coupled" appears and the solution is very slow.

I feel exactly that somewhere the residuals want to diverge, but they barely fluctuate around 1e-03.

Rob, are you saying that it is normal for the "Pressur Coupled" message to appear after each iteration?

What is the meaning of this message and why does it appear for me?

What is your suggestion?

what do I do ?

[Rob]

No, it is not normal. Please review comments on the mesh (not just cell) quality and try remeshing.