Hi all,
For a waste water processing plant I am trying to use MFiX to replicate the transport of sand inside concrete channels. Would this be suitable for the TFM solver? There is much less sand than water, and all of it will end up at the floor level. The geometry is full 3D, and so I expect it might be difficult to create a very fine bottom mesh, but perhaps it is doable. See sketch below. There are four inlets with a mix of water and sand, four outlets with only/mostly water, and also a separate pipe flow. So somewhat complex boundary setup.
Others may have more experience but to me this seems like a reasonable application of TFM. Try it and see!
Took some time, but looking into this case now. The geometry isn’t very complex, but I wonder if something might be wrong with the mesher still. Because even when heavily simplified, running single phase without any solids I can’t get it beyond the first timestep. Is the mesh the problem, or is this an issue with the graphics? Seems faces in corners might be missing. I have tried to change some of the mesher settings without much luck.
If the two regions on the first image are meant to be disconnected, you should have at least 2 to 3 cells in between. It looks like your mesh is too coarse.
No they should be connected. But is it a problem that corner cells seems to be missing? Or is it only a visual representation problem?
I think the cell corner not showing up is a visualization issue. Please attach your setup if you want someone to take a look.
Right now I’m not sure if I should make the mesh finer - if that’s the problem - or if it is the turbulence properties. So my first objective would be to get this geometry just to run in single phase. It’s difficult to determine what is actually causing the current problems. Can it be that I have four outlets set by 1 BC? Or that the outlet pressure BC does not work due to the hydrostatic pressure? Water is used as fluid.
After this I would add particles on the inlet, and observe how they are distributed.
BRA3Dtest_ny.mfx (7.7 KB)
channel_geometry1.stl (35.8 KB)
Hi Jeff, would you or any have a chance to give a feedback on this matter? Is the geometry too complex? Or might it be related to the specification of multiple pressure outlets?
Please try the attached. I tweaked some mesh tolerances, turned off k-epsilon because it was crashing, I changed the viscosity to that of water and increased the number of pressure correction iterations.
BRA3Dtest_ny.mfx (11.5 KB)
1 Like
Thanks Jeff. I appreciate that you had the time to look into this. I notice that you turned off reindexing, even though the domain consists of 70-80% of blocked cells. Are there still advantages to not enabling reindexing in this case, like the preconditioning?
The cell re-indexing is broken right now, we are looking into it.
I don’t get any error messages, and I got a sense that turning it on actually had an effect. But do you say that it does not really work, no matter if it is turned on or not? Or should it stay turned off to avoid something going wrong?
Enabling re-indexing often leads to a crash - a segmentation fault from indexing beyond array bounds. Search the forum for examples. We generally suggest keeping it off to avoid this problem. But if it’s making your case run faster without affecting the results, maybe it’s OK. We hope to have a fix for this soon but due to summer vacations, etc, we are a bit short-handed at the moment. Will follow up with more information when we have it.
Then I think I know what might cause this error - at least I recall getting some errors possibly related to this. I believe I have overcome them by not using the default “Autosize” values for the geometry dimensions, I always add a little in all directions. In my mind this makes sense, as cells are then not cut exactly at the same points as the geometry extents - I reckon this operations is “easier” for the algorithm when they’re a little inside the mesh cells instead. Though I haven’t looked inside the code, so I can’t really point to what’s causing this.
But I can at least say that I have noticed a real speedup when using reindexing, for geometries with a lot of blocked cells.
1 Like