strange solver behavior---Treatment of Rough Wall
 

It is well established that if the roughness height is much higher than the first node off the wall, the solver will blow up or at least gives a warning message. However, the thickness of the first element in my mesh is 0.01m and I imposed a roughness height 0.14m, which is 14 times the size of the first element. The solver runs smoothly and the result looks reasonable. But I AM NOT HAPPY!!! BECAUSE IT IS NOT WHAT I EXPECT. Something must be wrong. Somebody help me, please! Why I still can get a solution by defining such a large roughness height??????

Re: strange solver behavior---Treatment of Rough W
 

Hi,

I see no reason why making the roughness height greater than the first node should make the solver fail. It means the results will be inaccurate as the mathematical model used to model the effects of roughness is not valid but the solver should still converge to a solution.

Regards, Glenn Horrocks

Re: strange solver behavior---Treatment of Rough W
 

I always set my wall roughness height to 0.003ft, and first layer height of inflation to 0.001 inch, the solver never failed anyway. But How inaccuracy of the result would be? And is there any rule for this set up?

Thanks

Re: strange solver behavior---Treatment of Rough W
 

Hi,

How inaccurate? I have no idea. Do a benchmark study of flow in a pipe or flat plate with known roughness where you have good quality data to compare to (this should be easily available in the literature) and find out.

Glenn Horrocks

Does anyone know more about the behavior of CFX when the roughness is superior to the size of the first element, and when the SST model is used?

The behaviour is as described above. Convergence is fine, but the results are not accurate because the model is not valid in this situation.

I was looking for details, but I read the help files in between...

I do not understand why the model should not valid, given the rugosity treatment used by CFX. The rugosity effect is only a shift of the boundary layer velocity. Basically, the "under the roughness" nodes are extremely slowed down. And as the rugosity height used by CFX is the sand-grain equivalent rugosity, which is even higher, there's a good chance nodes will be under it. See the help which explains quite well...

Anyway, if someone has results showing the model is inaccurate, I'd be interested :)