CFD Online Discussion Forums

CFD Online Discussion Forums (
-   CFX (
-   -   Fatal Overflow when using RNG k-e (

Pascal February 5, 2008 05:12

Fatal Overflow when using RNG k-e

I am doing a study around and inside a front wheel of a car. Geometry is very complex but I succeeded to get a good mesh with prism and tri. Skewness below 0.9. I have already refined in the critical regions to obtain convergence. But I only get convergence for the k-e model. When I try to the RNG model, it stops at iteration #9 with fatal overflow although residuals seem to converge and the drag coefficient too. Physical timestep is 0.382s. Is the RNG model an appropriate model for such a complex flow? I found an old post (archive 2005) saying no but without any arguments! (It was just saying that it leads to false results with tetra mesh). Do you have any clue of how to solve this convergence problem?

Pascal February 5, 2008 06:39

Re: Fatal Overflow when using RNG k-e
Hi again.

It seems to work by dividing the physical time scale. (I divided it by 100). Wait and see now...

By does anyone has an opinion about RNG on tetra mesh? Below is the link of the old post saying it was not appropriated.


Glenn Horrocks February 5, 2008 17:41

Re: Fatal Overflow when using RNG k-e

In my experience the RNG model is very similar to the k-e model for results and stability. I know of no reason why RNG should be inherently less stable than k-e for tet grids. I think you have found the key, use a smaller timestep when starting up and ramp the timestep up as the simulation progresses.

Glenn Horrocks

All times are GMT -4. The time now is 16:37.