[DenysW]

p { margin-bottom: 0.25cm; line-height: 120%; } Ubuntu 16.04. openFoam v1612+. No parallel processing.


I started looking at this when I noticed that 2D openFoam was over-predicting settlement tank failure for conventional tanks at normal inlet-pipe velocities, compared with operating experience.So I checked whether 3D gave the same odd-looking velocity profiles within the stilling well.



Result: I’m getting a difference when I run the same physical geometry driftFluxFoam simulation in 2D vs 3D. In 2D the bulk flow moves inwards (-X direction) away from the stilling well wall at about half its depth, then goes down the central inlet pipe and eventually across the floor of the tank. This results in a major over-prediction of short-circuiting as a problem with this simple, traditional design of tank.
In 3D the flow only pulls away from the stilling-well wall at its base. There’s still a possibility of short-circuiting, but it appears greatly reduced.
So apparently the CFD doesn’t replicate its predictions between 2D and 3D simulations.
I attach screen dumps of the magnitude of U visualisations from paraFoam, and 2D and 3D tar.gz files with geometry from a 34 m diameter real-world tank designed in about 1995, simulated at its design maximum flows, not less than design feed sludge (0.0025 instead of 0.003). Apologies: the stilling well wall is hard to see in the 3D paraFoam case.
I've tried various detail differences in the 3D meshing in the inlet pipe and stilling well to make it as close as reasonable to the 2D mesh without any difference in outcomes.