rhoPisoFoam and time step issue: small dt INcreases INstability
 

Dear Foamers,

I would ask for your help to understand what I am doing stupid.
My experience, everybody should be able to quickly reproduce, start with the tutorial in $FOAM_TUTORIALS/compressible/rhoPisoFoam/les/pitzDaily (I am actually running version 1.7.1 from CentFOAM).

The solution obtained with the set-up already fixed in the tutorial gives non-physical results. Looking into the log it is immediately discovered that a too large Courant number is reached. A physically reasonable results is obtained changing the time step control to:
adjustTimeStep yes;
(you can stop the simulation very early, say endTime = 0.05).

Now, if you newly fix the time step to a constant value less than the minimum time step of the previous simulation, so that the max Courant number should remains within the stable range, the results appear to be newly non-physical. Actually velocity increases so much that Courant newly increases above the stability limit.

Therefore it seems that I am unable to increase the accuracy of the time discretization by decreasing the time step.

Please help me to elucidate this behaviour.

Kindest regards,

Franco

I ran the tutorial "as is" in OpenFOAM 1.7.x until t = 0.06s and I did not notice high values of Co, which stays under 0.5 from what I can see.

Should Co become very high also in the tutorial?

Best,

Hi Alberto !

thank you for your answer. I made a little mistake in my previous post as larger Courant numbers in the "as is" simulation are reached a little bit later. To report more precisely the issue, please look at the attached pictures.

Picture Courant.png compares the Courant number (Co) for three simulation with different choice of time step.
- v01 is a time step contrained to satisfy Co<= 0.5;
- v02 is a time step fixed to 5e-6 < of the minimum time step of v01;
- AsIs is the time step adopted by the simulation "as is".

Pictures U_0.05.png and Uv02_0.05.png compare the UX velocity component for the "as is" simulation and v02 simulation (both will show too large Courant numbers) at the same time = 0.05. While the latter is clearly unstable, the former looks reasonable. However by looking at the temperature fields, it is discovered that also the "asis" simulation is starting to have too large temperature variations that will blow up thereafter.
I made also some other simulations by further decreasing the fixed time step of the simulation but I only delayed the problem.

Therefore I concluded that the problem occurs taking the time step constant.

Best regards,

Franco