[saidc.]

Hi,

I'm trying to solve a laminar flow natural convection case that close-packed geometry. The cylinder's wall adiabatic. The mesh generated with this steps:

1. Generate sphere coordinates with desired porosity by DEM code (cemfDEM).
2. Create spheres and neck bridges for contact points approach with SALOME and fuse all together.
3. Generate 2D mesh with NETGEN 1D-2D algorithm with SALOME and export as a STL file (Images are attached).
4. Create background cylinder with blockMesh and create 3D grid with snappyHexMesh with desired features such as max-nonorthogonality < 55.

Necks image: https://ibb.co/xqqkbg9
2D mesh image: https://ibb.co/Nsv8n68

checkMesh results attached. For quick answer:

• 43 million cells (snapping start with 1 millon background cells)
• Max-nonorthogonality < 55 and average is 17
• Some (10-20) skewed faces.

buoyantPimpleFoam Solver settings are:

• 1e-6 p_rgh and 1e-8 U tolerance
• 1 nNonOrthogonalCorrectors
• 3 nCorrectors
• 1000 nOuterCorrectors
• relaxationFactors p_rgh --> 0.3 and others -->0.7

Initial conditions are:

• U (0 0 0)
• p_rgh (1 atm)
• T (fluid: 293K and spheres: 393K)

The problem is, pressure diverges and time step contuinity error is increases. After 50th inner iteration solver blows up. I tried change schemes (increased most robust to accurate ones), reducing tolerances, decreasing time step (1e-12 to 1e-21) but non of them helped. If I try to decreasing tolerances for control TSCE, pressure iterations reaches max and diverges.

Boundary conditions seem right to me but I know that if TSCE increases this is mostly related to BCs.

I'm open to suggestions. Files are attached.

Kind regards,
Said.