Convergence sensitivity to inlet turbulence quantities?
 

Hi all,

I wonder if anyone has experienced convergence problems connected to inlet boundary conditions for turbulence quantities in Fluent or CFX (or any CFD code).

Description of the issue:

I am trying to do a comparison between Fluent and CFX for a species mixing problem without reactions. The fluid is air with equal properties, which enters the domain through several inlets.

I started out from a converged steady state solution from CFX, in which the code had automatically computed the turbulence quantities at the inlet. Then I manually set constant values of k and epsilon at the inlets to agree with the values of this initial solution. I ran the simulation with these new (almost identical) boundary conditions and again it converged fine.

Then I went over to Fluent and set the same boundary conditions and used the same numerics as much as possible. When I ran the simulation it did not converge. Then I approximated the turbulence intensity using Eq. (7.3-1) in the Fluent 12 User Guide:

,

and the turbulence length scale:

.

I set these values as inlet values and ran the simulation in Fluent again, and the convergence was good.

However, when I set the same turbulence quantities in CFX I do not get a converged steady state solution. I ended up with running the CFX simulation in transient mode (for which the solution stabilizes after some initial (numerical?) transients) and comparing to the converged steady state Fluent simulation. Of course, it is time-consuming to run transient simulations so I would like to avoid that.



Does anyone have experience to share on this subject?


To clarify: with "converged solution", in addition to low residuals, I mean that the velocity and species fields are stable and are not changing with more iterations.