Learning Forum

[goktug.yilmaz]

Hello all,

I am currently working on a 3D analysis of a wing model, but I am facing convergence issues at a 6 degree angle of attack. The lift and drag coefficients exhibit unstable behavior, with fluctuations that make it difficult to obtain accurate results. In an attempt to resolve this issue, I modified the wing's trailing edge to make it blunt, which reduced the skewness value to below 0.9. However, the convergence problem still persists.

I tried to solve this problem by using Fluent Mesher to generate a polyhedral mesh. The orthogonal quality and skewness values of the mesh were quite good, with a minimum orthogonal value of 0.22 and a maximum skewness value of 0.72. However, when I switched to the solution using the "Switch to Solution" command, I obtained very high drag and lift coefficients. For instance, the lift coefficient increased to around 8000. This issue only occurs when I use the "Switch to Solution" command.

Could you please help me understand the root cause of these issues? Additionally, I would appreciate any suggestions or recommendations on how to address these problems.

Thank you in advance for your assistance.

[Federico Alzamora Previtali]

Hello, 

are you sure that your Reference values are consistent between your different model/simulations?

[goktug.yilmaz]

Thank you very much for your comment. I have checked and corrected my reference values as you suggested. Currently, I am not obtaining high lift and drag coefficients, which now seem reasonable. However, I am still experiencing convergence issues, particularly with lift exhibiting a fluctuating behavior at a 6-degree angle of attack. 

[Federico Alzamora Previtali]

Glad this helped!

Regarding convergence, how are your residuals looking? Are they still decreasing after 180 iterations?

You might need to run the simulation for more than that.

[goktug.yilmaz]

Residuals are also fluctuating. But now, I have tried to run the analysis with k-e model and Standard Wall Functions, this time it does not fluctuate. Residuals are decreasing, and I hope lift will converge a reasonable value, since it looks more stable now. Maybe I should focus on the boundary layer to be able to solve it with k-w SST.  Because previously, I used k-w SST model. Do you have any recommendations for y+ value? 1 is suggested, but it maybe hard for me to reach that value with my laptop. 

[Federico Alzamora Previtali]

k-w model uses a y+ insensitive wall treatment, but some of its advantages are lost if a wall function mesh is used.

We have a turbulence modeling course on the Ansys Innovation Courses space that summarizes this:
How to Model Near-Wall Turbulence in Ansys Fluent | Ansys Courses

[goktug.yilmaz]

Thank you very much for your answer. I have watched these videos. I understand that kw-sst is y+ insensitive and change the treatment according to mesh. I did not know this actually. 

By the way, when I refine the boundary layer by decreasing y+, polyhedral mesh exhibist skewness values that are greater than 1. Should I try to decrease skewness of these cells? I am unsure that quality check of polyhedral mesh is similar to tetra cells. Should I care about skewness for polyhedras? 

[Rob]

For poly cells use Ortho quality. The skew calculation gets a bit confused at times. 

Re the fluctuating values, plot a contour of pressure/wall shear run some iterations and repeat. Then compare the result. Then read up on boundary layer separation and stall. 

[goktug.yilmaz]

 

Thank you very much for your answer. I think a flow separation occurs since wall shear stress decreases a lot as you can see in the picture below. I also share my residuals to show how they fluctuate. Angle of attack is 10 degree. I could not solve the problem. It does not converge. Max y+ is equal to 5. 

 

 

 

 

 

 

 

 

 

[Rob]

You may need to increase the streamwise resolution near the leading & trailing edges to pick up the curvature more precisely: hard to say but the contour ought to be smoother. 

Well, if the separation point is moving you have answered your question. The flow isn't steady, so the residuals won't converge to text book levels. The question is whether the fluctuations are significant enough to justify the computational expense of switching to transient.