[kjetil]

Hello all CFX users,

I am evaluating someone else's work on a liquid pump, and I'm now investigating the numerics as the final results unfortunately do not reflect reality well. Y+ is generally good, and simulation was run with single precision, High resolution advection, and first order turbulence. Looking at some of the "velocity.beta" plots to see to what extent lower order upwind was used, I find there is a quite a mix of blending factors. Though the majority of beta values equal 1 (and second order UDS), there are I would say significant amounts of beta's lower than this. But how bad is this really? Would this give indications to improve the mesh or the setup, or is this just fine? For this specific simulation the dp/head does not converge, or at least fluctuates, while that is not found in test data at these flow rates.

[ghorrocks]

Why not run the simulation using hybrid differencing with a blend factor of 1.0? Then it is locked into using second order differencing everywhere. Then you can assess whether pure second order makes any difference to the small amounts of first order Hi-Res is giving you.