CFD Online Discussion Forums

CFD Online Discussion Forums (https://www.cfd-online.com/Forums/)
-   OpenFOAM (https://www.cfd-online.com/Forums/openfoam/)
-   -   simpleFoam crash -> How to solve (https://www.cfd-online.com/Forums/openfoam/88169-simplefoam-crash-how-solve.html)

tH3f0rC3 May 10, 2011 07:52
simpleFoam crash -> How to solve
 
2 Attachment(s)
Hi there,

I have tried to solve a case with simpleFoam.
Turbulence model: kepsilon, thus y+>30.

In the picture you can see yplus values of about >6000. I'm not quite sure if the value of yplus is limitted for kepsilon. I have only read greater than 30.
I would now suggest, to finer the mesh on the area of high values of yplus.

I have also uploaded another picture of pyFoamPlotWatcher. I still have tried to manipulate the value pf alpha (relaxation factors). I'm afraid I tried with no success. Does someone have an idea how to manipulate with success?
epsilon_max gets also a very strange values.

I'm happy for every answer.

Best Regards,
tH3f0rC3

tH3f0rC3 May 11, 2011 02:44
1 Attachment(s)
The new case I have tried to solve, the one with a finer mesh (yPlus max = 290) crashes also, but in a different way:
Please see the added picture

When I try to modify the alpha values (relaxation factors) I don't get better results.
Maybe someone of you has a better idea.

Best Regards,
tH3f0rC3

freemankofi May 11, 2011 13:11
it doesn't look like the crash is coming from "yplus" or the k-e model itself, rather it might be you're driving the flow "too fast". Try ff:

1. Consider, reducing the relaxation factors, especially the momentum terms.
2. If it persists, then try first with UDS (first order scheme) if you're not using it and then later change it to higher order to scheme.
3. try to review where which region has this abnormal feature and maybe, consider refining your mesh or something in that line....
4. if the above fails, then switch to different solver which allows the transient term. When using this method, change the starttime to be "0" and NOT "latesttime" since you're solving pseudo-transient problem. This method will allow you to use small timestep to drive the flow "slowly".

best wishes!

tH3f0rC3 May 11, 2011 14:56
Hi,
thanks for your answer.
Please see my notations below.

Best Regards,
tH3f0rC3

Quote:
Originally Posted by freemankofi (Post 307242)
it doesn't look like the crash is coming from "yplus" or the k-e model itself, rather it might be you're driving the flow "too fast". Try ff:

1. Consider, reducing the relaxation factors, especially the momentum terms.
OK!
2. If it persists, then try first with UDS (first order scheme) if you're not using it and then later change it to higher order to scheme.
Do you mean the setting in the fvShemes-file? The setting about gauss upwind,...?
3. try to review where which region has this abnormal feature and maybe, consider refining your mesh or something in that line....
OK!
4. if the above fails, then switch to different solver which allows the transient term. When using this method, change the starttime to be "0" and NOT "latesttime" since you're solving pseudo-transient problem. This method will allow you to use small timestep to drive the flow "slowly".
Can you list the solvers you mentioned? I only know steady state (where time step must be 1) or transient solvers (where one can set up the time step). So I think if I want to solve a steady state case, I have to use time step =1.
Thus, I would be happy if I thought wrong:D, because I also think that reducing the time step may solve the problem.

best wishes!

tH3f0rC3 May 12, 2011 07:07
I think I have now solved the problem.
I have used as boundary condition for inlet U:
type fixed Value

Now I am using
type
surfaceNormalFixedValue

and it works so far.

Best Regards,
tH3f0rC3


All times are GMT -4. The time now is 18:32.