Boundary layer simulation - convergence problems
 

Dear All,

I'm trying to simulate flow through a semiporous fence above a wall. This works OK when the fence is there, but I thought it would be valuable to check by removing the fence completely and computing the developing boundary layer (so essentially I'm computing a boundary layer on a flat plate - I would think something I should be able to get to work!). Under these circumstances I can't get Fluent to converge correctly - the continuity residual levels levels out at 1e-2, and this eventually (I think) stops the rest of the residuals converging.

Some details : the Reynolds number of the flow at the location I'm interested in, based on the distance along the plate, is 20,000. I'm using the k-w model with the Transitional Flows option enabled (so its essentially using a Low Reynolds Number formulation). The mesh is rectangular, quads, with near-wall boundary layer refining to give y+<4. Inlet turbulence is 0.4% of the free stream kinetic energy. I'm using the defaults on the solver - SIMPLE, 1st order upwind etc, although I am starting to try altering these to see what they might do.

Any suggestions or hints as to what might be worth trying here would be great! Thanks in advance.

Gavin

Re: Boundary layer simulation - convergence proble
 

Hi,

I don't have experience with fluent but from what you are writing about too high continuity residual levels seems to be an issue of the boundary conditons at inflow and outflow. If you specify fixed inflow, the outflow condition should be appropriate (don't know what's available in Fluent). Usually one fixes the pressure. One could also try periodic boundary conditions at in- and outflow with fixed v at some height to check this.

regards

Volker

Re: Boundary layer simulation - convergence proble
 

I'm using a fixed velocity inlet/fixed pressure outlet, with symmetry on the top and a wall boundary at the bottom - pretty standard really. Moreover, when I put the obstacle in the flow the solution converges very nicely.

Looking through the Fluent manual, they suggest that if the initial conditions are fairly accurate, the initial absolute residual in the continuity equation may be quite small and so the relative residual may remain quite high. Does that sound at all plausible? (I'm always rather suspicious of commercial CFD companies trying to justify why their codes aren't converging properly...). Certainly the continuity error is low (1e-8) and uniform across the entire flow.

Gavin

Re: Boundary layer simulation - convergence proble
 

Try double precision, that should answer the question of residual. I don't think this is the default.