CFD Online Discussion Forums

CFD Online Discussion Forums (http://www.cfd-online.com/Forums/)
-   FLUENT (http://www.cfd-online.com/Forums/fluent/)
-   -   Convergence (http://www.cfd-online.com/Forums/fluent/41849-convergence.html)

Ramine July 26, 2006 07:15

Convergence
 
Hi all,

I've checked in the forum for previous answers to my question and I'm still not sure...

Here's the problem: I'm modelling 3D hydrofoil (unstructured grid of 900k cells - no BL yet...) at low-moderate Re in fluent 6.2.16 and after about 500 iterations, the convergence stops decreasing and continuity stabilizes around 10^-4. Also, the other plots I am tracking stabilize or oscillate around a value.

I would say my case is converged but I have doubts as the residuals do not go down...

I am using laminar solver, second order discretization scheme for the momentum, standard one for the pressure and SIMPLEC with skewness correction of 1 for now...

I'll refine my grid around the foil when my case will be converged in order to use a turbulence model later.

Any thoughts of what I'm doing wrong?

Cheers,

Ramine

Razvan July 27, 2006 14:44

Re: Convergence
 
As a very experienced user of FLUENT, I'm telling you that continuity residual convergence at 10e-4 is perfectly possible. I can also say that I encountered cases where due to boundary condition settings, geometry of the flow domain and solution initialisation technique, I obtained convergence of the continuity residual at 10e-1 !!!

You have to understand the fact that residual drop is very much dependant especially on the solution initialisation, e.g. if the initialised flow field is very close to the converged flow field, naturally the residual drop for continuity will be small.

Residual oscillations could be due to the grid. Unstructured grids inevitably contain highly skewed cells. If these cells are placed in important regions of the flow domain, with high gradients (like near the wall of the hydrofoil), then numerical diffusion and dissipation generated by these particular cells could determine residual oscillations. Another reason might be the well known instability of the laminar solver, which in particular cases could force one to use the unsteady solver for obtaining a converged solution. Not the case here...

Razvan

Ramine July 27, 2006 17:01

Thanks !
 
Ok, thanks very much Razvan!

Regards,

Ramine

Razvan July 28, 2006 01:28

Re: Thanks !
 
You're welcome!

Razvan


All times are GMT -4. The time now is 06:46.