[wildfire230]

Hi all,

I am working on several simulations in which the sizes and volumes of cells range by orders of magnitude. One of these cases is basically a microfluidic channel that opens into a much larger main channel, and the dynamics within the pore and in the immediate vicinity of the junction with the main channel are the main points of interest.

In my current simulation, I have these results from checkMesh:

***High aspect ratio cells found, Max aspect ratio: 1006.08528449, number of cells 800
<<Writing 800 cells with high aspect ratio to set highAspectRatioCells
Minimum face area = 6.25e-12. Maximum face area = 5.45656661211e-06. Face area magnitudes OK.
Min volume = 8.20300604974e-17. Max volume = 4.02026660785e-09. Total volume = 3.005125e-06. Cell volumes OK.

You can see that I have both very high aspect ratio cells, and also the maximum and minimum cell volumes range by 8 orders of magnitude. The flow is low Reynolds number and steady, but I am not getting nice convergence with simpleFoam, and my best guess is that it is because of the mesh quality. So I have two questions:

1. Do high aspect ratio cells prevent nice convergence with simpleFoam?

2. Does having cells with many orders of magnitude different volumes affect convergence with simpleFoam?

Thanks in advance!

[piu58]

Cells with high aspect ratios are o.k. or even necessary, if the flow / pressure gradient in one direction is much higher than in the other. Often you have this in the near of surfaces. In region whee the flow spreads and disposes the flow energy to the whole volume such cells are undesirable.

[wildfire230]

Hi, thanks for the response. That makes sense. Is there any limitation on the range of cell sizes that you can have? For example, suppose I have a 50 micron diameter channel that opens into 1 cm diameter main channel. Then of course the cells in the main channel will be much larger than those in the small channel. Does that affect convergence? I know that simpleFoam uses some kind of relative error for measuring convergence, and I think I heard somewhere that the large cells might dominate the calculation, so that the relative errors could be small, although the results in the small cells are highly inaccurate. Do you have any experience or thoughts with those issues?

Thanks again!

[piu58]

50 µm against 1 cm looks large. But it may be adequate to your case.

I recommend th following:
Let the analysis run until you get a rather stable state. It doesn't has to be an equilibrium, but avoid the very first time steps for this.
I always use the time control by maxCo. If you do so, you see what time step you get. If may change something, e.g. changing the 1 cm coordinate to 3 mm. If you get a larger time step then, you move in the right direction. If the time step doesn't change or get even smaller, than you did the wrong thing.

Mesh analysis is important in every case. In such situations you have additional to check the mesh resolution between the directions.