Divergence with LU-SGS
 

Hey everyone!

I am trying to verify my implementation of the LU-SGS scheme and I am having some issues. First, I am checking the free-stream capture property of the code, which apparently is not working.

I used a rectangular box with uniform spacing. As soon as I have a small perturbation at the boundaries, it is slowly amplified until a pattern as shown in the attached figure starts to appear. Has anyone already seen something like this before? Where could it come from?? I believe that my calculation of the residual is ok since I previously used it successfully with my explicit solver. I am using central differencing, so that the jacobian is computed analytically.

Ant idea would be greatly welcome!!

Thank you for your help!

Joachim

and here is the picture...

If I use a low CFL number (~0.1), the free-stream capture works just fine and the full simulation converges (checked with Blasius and a naca0012 airfoil). I guess that confirms that my residual is ok but there is an issue with my implicit operator?

How could I identify the precise source of the bug??

Thanks!!

Joachim

without detalis about the formulation and the code is not possible to help you...