strange observations: 1

kenn

Hi, I wrote a 3D parallel FEM (with all kinds of stabilized terms) for incompresible transient N-S. when I run benchmark problems I always have this kind of trouble.

for example, in cavity flow, if USE 10x10 mesh, i can get a result. if use 20x20, result becomes better. using 40x40, it becomes even better. howvere, if refine mesh again and use 80x80, guess what, the result is totally rediculous.

what ever in pain Galerkin formulation or with all additional terms which stabilize pressure and convection terms, this kind of phenomena always heppen, sooner or later.

I am pretty frustrated and wonder if some one could shed a light?