[GerhardHolzinger]

Hi,

the commit message [1] for the introduction of the face-based momentum equation states, that the face-based momentum formulation allows for:

improving stability and allowing larger time-steps

I have attached a rather simple bubble column case, which I have been playing around. I have been run able to run the simulation for a maxCo = 0.6 setting in controlDict, for both the face-based formulation as well as the cell-based formulation (faceMomentum true and false in fvSolution).

For a setting of maxCo=0.75, the simulation crashes within the first 5 seconds, for both formulations.

Using the face-based momentum equation, however, improves the solution in a way, that it does not show some sort-of checkerboard-like artifaces, see the attached PNGs.


My working theory is that the two-fluid Euler-Euler model is not happy with Courant numbers larger than 0.5. Single-phase solvers, on the other hand, can be run with maxCo settings up into the low two-digits, e.g. I ran a vortex shedding case up to maxCo=20 with a reasonable accuracy of the vortex shedding frequency.

So now the challenge: Does anybody find a case-setup, which allows for larger time steps respectively Courant numbers than the ones I achieved so far.

References:

[1] https://github.com/OpenFOAM/OpenFOAM...777bb10a3e1dda


Rules:

• Post your results along with the case setup as archive.

• All posted cases should be able to be run with an Allrun script, i.e. all necessary steps must be contained in the Allrun script.

[evamaria]

Hi,


I am not sure if my question/finding fits here, but I thought it is the most suitable thread.

I recognized, that the solution using reactingTwoPhaseEulerFoam in OpenFOAM-7 as well as multiphaseEulerFoam OpenFOAM-8 strongly depends on the maxCo number, even if maxCo is always well below 0.5 (see pictures).



Do you have any ideas why this is happening? Is it okay or is there a problem with the solution algorithm? (or maybe only with my case-setup?)

My idea was, that the differences come from the MULES algorithm. Possibly, different time steps lead to a different blending of high-order and low-order schemes and therefore to different "diffusivities" of the solution.

I attached the case setups for reactingTwoPhaseEulerFoam (OF7) and multiphaseEulerFoam (OF8).