convergence criterion
Hi I am wondering if someone can tell me what factors influence the atteinable convergence criterion. I am studying the turbulent flow over a descending step and am having a very hard time getting Fluent to converge. With a KEpsilon model, and the Enhanced Wall Treatment, the results with a conv. criterion at 10^4 are not acceptable. With 10^5, it is better but not yet the right answer. The recirculation zone at the foot of the step is much too small. I even tried to apply a laminar model and still, no convergence with the convergence criterion at 10^4. All these, with the same mesh.
Can anyone help?? Also, is it normal for k to be more or less 100 times smaller than Epsilon? Thank you in advance! 
Re: convergence criterion
Hi, Did you try with the underrelaxation coefficients? And did you try changing the discretization scheme?
Play a bit with them to see what happens. Take care. 
Re: convergence criterion
Hi, Well I am starting to play with the underrelaxation coefficients (which is what my professor suggested), but I'm looking for some guiding principles. I tried several discretization schemes and obtain the exact same results every time. Any idea why my laminar results aren't acceptable either? Could it be that I shouldn't be using a pressure outlet (of 0)? My inlet is a constant velocity inlet... Thanks again Dominique

Re: convergence criterion
Is your inlet far enough away from the step? Maybe the velocity profile needs some more time to develop.
RoM 
Re: convergence criterion
Well, for now I am using a constant velocity at the entrance. Thanks anyways, any other ideas? Dominique

Re: convergence criterion
You can use a constant velocity if you dont have a profile but you have to make shure that the inlet is far enought away from the step so that the velocity profile has some time to develop. I think the distance should be at least 10*H where H is the channel height.
RoM 
All times are GMT 4. The time now is 23:43. 