[cbooks77]

Hello,

I am simulating a mixing tank using the MRF approach to handle the impeller region. It is well known that in highly swirled flows the turbulence cannot be treated as isotropic hence it is most appropriate to use the RSM Stress Omega model for highest accuracy.

I am currently validating my model and attempting to match the manufacturer specified dimensionless power number for the impeller. I have performed grid dependency studies and for each grid density I have studied the RKE, kw-sst, and RSM Stress Omega.

The RKE and KW-SST predict torque around 11.5 Nm and volume average turbulent dissipation of around 0.01.

The RSM Stress Omega with 1st order turbulence resolution predicts torque of 13 Nm and turbulent dissipation of around 0.005.

The RSM Stress Omega with 2nd order turbulence resolution predicts torque of around 12 Nm and turbulence dissipation of around 0.01.

The problem is that the RSM 1st order turbulence discretization predicts the torque correctly at 13 Nm while underpredicting turbulent dissipation. But, when I switch to the 2nd order resolution the torque value drops to around 12 Nm and the power number is around 12% error while the turbulent dissipation falls in line with the other 2 eqn models.

Interestingly when I use the sliding mesh approach with 2nd order turbulence resolution the torque goes back up to 13 Nm.

Can anybody provide an insight into what could be occurring here? I am wondering if this is just an issue with the MRF approach and its inherent weaknesses. I will note that I have made the MRF region large enough to capture the vortex shedding off blade tips and there are between 10-20 nodes along the blade thickness. 2nd order momentum resolution is used in all cases. I am just confused as to how the 1st and 2nd order turbulence resolution could have such a significant affect on the torque predictions of my impeller.

[LoGaL]

Is the second order case properly converged?

[cbooks77]

Originally Posted by LoGaL

Is the second order case properly converged?

Hi Lorenzo,

Yes all of my flow variables stabilize to a periodic steady state and residuals are very stable. Poly-hexcore Mesh has minimum orthogonal quality of 0.5.