Stability issues when introducing CSG Geometry in MFIX-Exa

MFIX-Exa
1

Hi everyone,

I am relatively new to MFIX-Exa and have been progressing through increasingly complex cases. Recently, I’ve hit a bottleneck with simulation stability.

My CFD-DEM simulations run perfectly fine (stable and converging) when using a simple box/domain without complex geometry. However, as soon as I introduce CSG geometry, the simulation becomes unstable or diverges early on.

I suspect the “small cell” problem where fluid cells are cut by the CSG boundary.

Questions:

What are the best practices for mesh refinement around CSG boundaries to prevent Divergence?

Should I be looking at the mfix.fixed_dt or letting the adaptive time-stepping handle the reduction?

Any advice on the common gotchas when moving from box-simulations to CSG-defined boundaries would be greatly appreciated!

Case:

geometry.csg (456 Bytes)

inputs.txt (8.1 KB)

Outputs:

mfix.out.txt (336.5 KB)

Backtrace.3.txt (3.3 KB)

Backtrace.2.txt (3.3 KB)

Backtrace.1.txt (3.9 KB)

Backtrace.0.txt (3.3 KB)

2

Update: Stability issues resolved by addressing small cut cells.

Settings borrowed from @YupengXu

mfix.grid_pruning = true

amr.refine_grid_layout = false

eb2.cover_multiple_cuts = true

eb2.maxiter = 1024

eb2.small_volfrac = 0.1

2 Likes