I do a quasi-3D airfoil simlation, and I had generate one element(2mm) 3D grid. when i use SST model to calculate, I always get the follow error message after several steps of iteration.

ERROR #004100018 has occurred in subroutine FINMES.

Message: | Fatal overflow in linear solver. I found that the "wallscale" variable can not converge. but other variable, such as velocity mass and k and W, can converge. I don't know why. Does it have something with the grid quality? The largest width-height ratio of my gird may be 400. That is because I have made a very fine grid near the wall (0.04mm),so this will contribute to large width-height ratio in some position.

can anyone give me some suggestion? thanks a lot .

Wallscale is a simple diffusion equation. If it is failing, it is probably due to poor mesh. Large aspect ratio's should not be a problem (and 400 isn't very high), but if they are the problem would go away by running double precision. I recommend opening your model in Post and looking for elements with very small edge angles.

-Robin

Hi, Robin

Thank you for give me the recommendation. What I want to say is that my mesh was generated by Gambit2.1. And I have checked the quality of the mesh. except for the aspect ratio, the equiangle skew of the mesh is smaller than 0.1. So I think the quality of the mesh should be very good,except for the large width-length ratio.(it is not aspect ratio,maybe I didn't describe clearly before.)That is to say, some grid should be very long and very thin, but the angle of close to 90 degree.

Hi Robin

Just now I have changed the aspect ratio of my mesh. Because I don't know the original value of aspect ratio exactly, so what I can show you is that I decreased the aspect ratio 8 times. and than it can work. I think the problem should be contribute to the poor aspect ratio.

Could you tell the normal value of aspect ratio?

-Fanzhong Meng

Hi Robin,

I found your comment a bit confusing. I remember reading somewhere in the manual that solver may have problem if edge length ratio > 100.

Is 400 far beyond the limit specified in manual? Is there any significant improvement in Version 10 that allow edge length ratio to go beyond 100?