Convergence of the polydispersity drag models in MFIX-PIC

Dear MFIX Contributors,
I have a simple case using MFIX-PIC method for polydispersity calculation.When I used the basic drag models like Wen-Yu and Gidaspow drag models, the calculation is converges. However, when I used the polydispersity drag models, such as BVK, GIDASPOW_PCF and GIDASPOW_BLEND_PCF drag models always showed the following:


This is my case:fluid_bed_dem_2d333333.mfx (15.2 KB)
I hope to get your help, thank you!

I can reproduce this issue with the PIC solver, while the polydisperse drag works fine with DEM solver.

I have created a ticket: #1005

@gaoxi Thank you for your answer. The polydisperse drag is fine for DEM solver and MFIX-PIC single-component calculation. I look forward to this problem being solved.

I am surprised that the DEM solver is available to use with the polydisperse drag models in MFIX. In the current implementation, they were derived for and should only be applied to TFM (i.e. Eulerian solids).

Also, it is my impression that polydisperse drag models, in general, make the most sense when a system has thousands of particles per cell, and I doubt that is the case for any DEM simulation, and given your input sheet, not the one you are running.

As for PIC, if you want Gidaspow-Blend drag, please do not apply the _PCF version (or any of the other _PCF drag models available in MFIX). The program will definitely fault because solids fraction by phase is not tracked through a full simulation, as it is unnecessary for the solids model.

For MFIX-PIC, the drag models are employed through total solids fraction only, and by default the diameter of phase 1, hence why the _PCF models don’t make sense. You certainly CAN track solids fraction by phase, if you want to, but I would suggest doing so through a UDF where you might apply a more Lagrange-friendly polydisperse drag function.

Finally, I note in your input sheet that you have applied PIC to an essentially 2D simulation. You can imitate 2D, that is true, but you need at least THREE (3) cells in the K-direction for the PIC calculations to be meaningful. The 3 minimum is required for the interpolation method.

Meanwhile, thank you for pointing out this problem with the current _PCF drag model selections in MFIX. We need to employ a preventive measure to keep users from selecting this option in the future.

2 Likes

@maclarke Thank you for your answer. I have a deeper understanding of the scope of application of the polydisperse drag models in MFIX.