Diverging timestep due to turbulence residuals

I have been trying all day to get a 3D TFM simulation to run, but it keeps failing on what seems to be turbulence residuals (and a few times p0). I have run this simulation - or one very similar - a lot of times in 2D already, but this time it is unusually difficult to make it run properly. I have tried numerical schemes (both default and SMART), initial time step, checking initial conditions for pressure and turbulence, and increasing meash small cell tolerance and snap tolerance. I have a feeling that the mesh still might be a problem, or the inlet condition, as I read from the terminal that max allowed velocity is also reached when the timestep plummets. Does somebody have suggestions to what to troubleshoot next, or what might be the possible problem?

And is it so that the cells marked as “small_cell” is actually removed when the solver starts?


Was it this message?

mfix - WARNING - velocity higher than maximum allowed velocity:   25.000    (to change this adjust the scale factor MAX_INLET_VEL_FAC)

That one should be getting detected by the GUI and contribute to the “Warnings” count but it doesn’t - we’ll fix this in the next release.

Have you tried changing MAX_INLET_VEL_FAC ?


It was that message, except max allowed velocity was something like 50000m/s. I am not expecting anything larger than 3m/s though. I’m using a mass inflow, not velocity, could this be the problem? I am not allowed to specify a pure velocity inlet though, but can give optional velocity components below.

I actually made it go longer by fixing the time step, setting a hydrostatic pressure IC, and disabling detect stall. Perhaps I should look further into improving the mesh?

Improving the mesh is always a good idea!

Even after simplifying the geometry and putting quite some effort into a good mesh, the divergence knocks in at about the same time, around 0.5s, and starting to complain about negative gas density.

Another strange thing is that the outlet monitor reports significant amount of particles being pulled out the top, even from the start. This shouldn’t be possible. This goes for all phases. The air phase remains constant up to 0.5s, but the solid phases cross over to the negative side. I have not seen this happen in the 2D simulations, which otherwise are very similar.

I have rerun this with a simple cylindrical geometry, and still see the same “unphysical” behaviour where particles are at the outlet initially, in addition to odd arithmetic errors linked to file saving. This is a bit of a headache, because none of these problems were present in the quite similar 2D simulations I ran until now, and seems not to be a problem in the 3D TFM tutorial either.