CFD Online Discussion Forums (http://www.cfd-online.com/Forums/)
-   FLUENT (http://www.cfd-online.com/Forums/fluent/)
-   -   Highly Skewed Cells (http://www.cfd-online.com/Forums/fluent/78390-highly-skewed-cells.html)

 chrisoturner July 20, 2010 10:22

Highly Skewed Cells

Hi!

I am running a simulation that has approximately 500,000 cells. It is a hybrid mesh of both hex and tet cells. The tet cells are being used as this is an optimisation problem where different geometries are being modelled and compared-the tet cells seemed to offer the most consistent cell size and count for different geometries compared to the hex options.

In my model, I have approximately 0.01% highly skewed cells (above 0.97). My model still converges and seems to give reasonable results so I was wondering if these skewed cells matter? There are approximately 40 cells that are highly skewed in the entire model and none of them are clumped together, it seems to be one cell here and one cell there.

Does anyone know of a way of getting rid of single highly skewed cells within fluent /Gambit?

Thanks!

Chris

:o

 -mAx- July 21, 2010 01:03

In Gambit, if you are sure that the highly skewed cells don't come from your geometry (small angle, small edge, etc...) then you can try modifying slightly the cell's size (or Size Function's parameters).
That's my workaround

 kdrbrk July 21, 2010 04:38

can we just change the mesh size of skewed cells, or must we remesh the whole domain?

 -mAx- July 21, 2010 05:07

you must remesh all the volume concerned.

As another workaround, if the highly skewed cells are direct connected to a surface, you can try to correct the skewness of the cell by moving the node (belonging to the cell) on the surface (--> Face/Mesh/Move Face Nodes)

 kdrbrk July 21, 2010 09:56

my highly skewed cells are generally on(2D) or next to(3D) surface. So these recommendation will be very useful. Thanks Max :)

 chrisoturner July 21, 2010 10:08

Thanks for the info Max. In your experience, if there are very little skewed cells (0.01%) and the CFD matches experimental data, is there any need to try and eliminate the skewed cells?

I just think it may take a long time for not much improvement in result accuracy.

Chris :)

 -mAx- July 22, 2010 01:19

Quote:
 Originally Posted by chrisoturner (Post 268346) Thanks for the info Max. In your experience, if there are very little skewed cells (0.01%) and the CFD matches experimental data, is there any need to try and eliminate the skewed cells? I just think it may take a long time for not much improvement in result accuracy. Chris :)
In my experience, I can say that highly skewed elements can affect the convergence with drops in the residuals (but not always).
But the question is: Are skewed cells in a specific region where you need to catch something?
At the other hand you reached results in agreement with experimental data, so....
I already gave up trying to fix some highly skewed cells, especially after one day of attempts, and I didn't encounter any convergence issue. :rolleyes:
Nevertheless I try to keep max skewness below 0.9 (for unstructured grids)

 chrisoturner July 22, 2010 06:43

Thanks for the advice MAX! Once again, you have been very helpful!

Chris:D

 All times are GMT -4. The time now is 21:19.