[jdp810]

Hello all,

I am getting the skin friction of a profile but, even though the values are close to what I am expecting, the skin friction oscillates throughout the profile as you see in the attached picture. I also attach the config file I am using. It must be something about the MUSCL scheme, but if I deactivate it (convective part) the solution is no longer correct.

How can I solve this?

Thank you,
Jose

[pcg]

Try using the Venkatakrishnan-Wang limiter instead and experiment with the limiter constant a little.
Is the problem fully converged? Or do you have some oscillations in the residuals?

[jdp810]

The residuals end up oscillating. However the coefficients are converged. I have already tried changing the constant of the Venkatakrishnan limiter to a higher one but didn't have any noticeable change.

I'll try with the Venkatakrishnan-Wang next week, hope it solves it.

[pcg]

That does not look converged enough.
I am not familiar with the incompressible solver and with 2nd order turbulence, most folks run it 1st order (even in other codes).
My experience with the compressible solver is that high CFL can sometimes stagnate convergence as the linear solver starts producing gibberish, this is harder to happen in 2D but... you can add LINSOL to the HISTORY_OUTPUT to see what convergence you are getting there.
If the grid is not completely unstructured multigrid may help a lot, try 2-3 levels, V-cycle with (1,1,1,1) pre-smoothing, all the other (0,0,0,0) and the damping factors to 0.7.

[jdp810]

So after trying with the VW limiter with a coefficient of 0.5 I got much better convergence as you can see, even the pressure coefficient is way smoother, thanks. The linear solver is also converged, it always finishes before the maximum iterations (I set 20) per step at around 1e-9 residual from the LINSOL output.

However, the skin friction has still the sawtooth shape. I run it at different alpha, but it appears at all alphas...

[bigfootedrockmidget]

Residuals are still going down, could you run it for another 5000 iterations and compare the results? Also, what is your y+ value? is it around 1 everywhere?

[jdp810]

It looks converged now, but the skin friction profile is not smooth yet... The y plus is less than 1 for the entire profile except for the leading edge, but still, it is very low, I don't think that it is causing that behaviour.

[pcg]

I would try first order turbulence MUSCL_TURB= NO
You can also try lowering the CFL to see if it clears things a little, I find it odd that 20 iterations is enough to get 1e-9 residual.

[jdp810]

Originally Posted by pcg

I would try first order turbulence MUSCL_TURB= NO
You can also try lowering the CFL to see if it clears things a little, I find it odd that 20 iterations is enough to get 1e-9 residual.

Sorry, I didn't mention that, when I changed to VW I already turned off second order turbulence and increased the linear iterations to 50. You can see the first 2000 iterations history in the zip file. Do you find something weird?