K-w model again !!!
 

Dear Friends,

I am implementing k-w model and am faced iwth a serious problem, something I had also described in an earlier post a week back. In fact, on grids with large AR cells, typical of most viscous flows, my solution gets corrupted. What is observed is that the solution develops well, and reaches the "expected" steady state(which is known because we are trying out a standard problem on a given grid, and additionally solution from an algebraic model is available, which can give a reasonable qualitative comparison) by around 2 to 3 decades of residue fall, but does not sustain this state, eventually leading to a non--physical solution and code blow-up. Jonas remarked that it could be because of near--wall problem and treatment of the Omega b.c., which I now am using surface roughness based BC (by Wilcox). Interstingly, the source of trouble seems to be at the trailing edge for the airfoil problems I am trying. I see that the skin friction curve shrinks( ie the predictions in the rear of the airfoil inc. TE decreases) appreciably at the rear portions. I believe this is due to a corrupted du/dy, since turbulent viscosity is zero at the wall. I am using a linearity--preserving procedure(Green Gauss reconstruction + Pseudo--Laplacian) for the gradients, but I am not able to really understand what is the cause of this corruption from the right result. I would greatly appreciate if anyone who has implemented the model and validated it for flows with grids with considerably high AR ( since low AR causes no problems for me, as of now) gives their views/opinions/experiences. Any comments and suggestions are most welcome.

Thanks in advance,

Regards,

Ganesh