CFD Online Discussion Forums

CFD Online Discussion Forums (
-   FLUENT (
-   -   Convergence problems - please help (

M Liddell February 8, 2005 14:08

Convergence problems - please help
I'm running an analysis on a 2d sail section with mast using k-omega sst.

The grid was created in Gambit in cm and I changed the 'grid was created in' to cm in fluent and hit scale. The grid extents are then correct.

The continuity, and velocities will not drop below 1e-03 no matter what I do with the under relaxation. However if I do a run WITHOUT scaling the grid it does converge to 1e-06 in about 1800 iterations. PISO diverges massively while simplec is no different to simple. I've tried first and second order for momentum and turb. KE to no avail.

This problem is driving me nuts, I've read everything I can find and tried increasing the mesh density in high gradient areas but I just get the same. Any help would be massivly appreciated!

Jason February 8, 2005 14:24

Re: Convergence problems - please help
Your convergence isn't only a function of the residuals. You should monitor whatever it is you're trying to get out of your model (lift, drag, pressure at a location, etc...) and use that as convergence criteria. Since the residuals are normalized based on the initial residuals, if the initial guess is close, some of the residuals won't drop as much, typically continuity. Often you can call your model converged without the residuals fully reaching the convergence criteria set in Fluent (this value is a typical convergence criteria based on past experiences of the FLUENT team... it does not in anyway guaruntee convergence and you should always monitor the results you're going to report to show that they are converged).

Another thing that's mentioned a lot in past posts is using the 2ddp solver instead of the 2d (the dp stands for double precision... since residuals are a way of looking at numerical error, if you double the precision of your numbers, you can dramatically reduce the numberical error... I don't think this is a required step, and I haven't seen much of a change in anything but the residuals when switching to this solver, but in some cases I'm sure there are models where it has an impact... give it a try and see how much your forces change).

Hope this helps, and goodluck, Jason

M Liddell February 8, 2005 17:21

Re: Convergence problems - please help
Thanks for your response, it's cleared up a lot. My Cl and Cd values are pretty converged so I think it's all ok.

zxaar February 8, 2005 20:06

Re: Convergence problems - please help
i have observed one thing that sometimes, if we define interface and then scale the grid.. the grid check just after it fails, so for that reason i usually scale first and then define the grid interface .... so if you are defining grid interfaces ..there is a possiblity that this thing might have happened to you and when you run with scaled the grid is bad and solution diverges ...if this is the case .. check it

All times are GMT -4. The time now is 14:58.