The accuracy issue of CFDEMSolverIB on drag calculation

Hi CFDEM folks and developers,

I have been playing around the fictitious domain method (FDM) to calculate the drag on a spherical particle for some time. It seems that there is a accuracy issue in terms of the drag calculation even with very fine mesh (error up to 30%).

A bench-marking test was performed to simulate the settling velocity of a single spherical particle falling freely in water. The simulation conditions are:

FLuid domain: 0.1m x 0.1m x 0.6m (L/W/H) filled with water (nu=1e-6, rho=1000)
Particle: a single spherical particle of 0.01m radius, initial position: (0.05, 0.05, 0.52), initial Velocity (0, 0, -1.5), density 2500kg/m3

The analytical settling velocity is calculated with the following correlation for Re < 2e5:
(Brown, P. P. & Lawler, D. F. (2003). Sphere drag and settling velocity revisited. Journal of Environmental Engineering, ASCE, 129, 222–231)
--- Rep = (rho_f*dP/mu_f) * U
--- Cd = 24/Rep (1 + 0.15 Rep^0.681) + 0.407 / (1 + 8710*Rep^-1)
--- Fd = 0.5* Cd * A * rho_f *|Uf- Up| *(Uf- Up)

The calculated settling velocity is c.a 0.94 m/s with Re = 2e4. However, the simulation results of CFDEMSolverIB with different mesh resolution still have relatively large gap to the analytical solution.

Settling velocity = 0.68 --- Coarse mesh: 10x10x60; dynamicMesh settings: nBufferLayers 2; maxRefinement 2;
Settling velocity = 0.74 --- Intermediate mesh: 20x20x120; (same)
Settling velocity = 0.79 --- Fine mesh: 40x40x240 (same)

It is obvious that the finer mesh is the better results we get. For the last case, the cells around the particle-fluid interFace are extremely small compared to the particle size, but the simulation result is still not satisfactory (error 16%). Plus, due to the refinement process around the particle surface, the increase of cell amount is significant (5k to 80k increase of sub-cells). That being said, the resolved CFD-DEM coupling seems not practical in most cases, given the fact that the accuracy issue and the fine mesh size required.

Maybe the settling particle test s not enough to benchmark the resolved method, I will give a try with a fixed particle in fluid and compare the drag between pure CFD and the CFDEMSolverIB simulation.

Any advice or comments?

Best regards,

David

___________________________________
oops: I accidentally put some keywords in the "Taxonomy upgrade extras", but resulting in few forums topics :(, can you please delete these, admin? thx