[jseamon1]

Hello All!

I am looking at pressure-driven flow through small micro-tubes (<500 micrometers in diameter). As part of my model verification, I was performing a mesh refinement study and looking at how the mass flow rate changes with increased mesh refinement. The first figure (stationaryFlow.png) below is what I was getting for each of the 5 runs (test0: Least Refined Mesh, test4: Most Refined Mesh). The mass flow rate differs only slightly across the tests, but the fact that it generally decreased doesn't really make sense as the max velocity of the cross-sectional velocity profile increases with the mesh refinement (see the second figure, stillProf.png). As a check, I decided to integrate the cross-sectional velocity to calculate the flow rate and was surprised to find that there was quite a bit of disagreement between what STAR-CCM+ was giving as the flow rate and what I calculated through integration (this is seen in the third figure, compareMassFlow.png).

So, my question is why is could anyone explain how mass flow rate is calculated in STAR-CCM+ and why my results would differ so much from those given by STAR-CCM+? Thank you in advance for your help!

[fluid23]

I believe that it sums the flowrate of each cell at the centroid. If you are creating a derived part > plane section to capture flowrate away from a boundary, then it will have to interpolate from the nearest cell center to the plane. However, it is a direct sum of the flow in each cell. If you have a boundary condition enforcing a particular flowrate, then you won’t see that report change too drastically with a refined mesh.

It looks like the mass flow report is staying roughly constant (as you would expect), but the integrated value is changing because your cells are shrinking. Think of a Riemann sum where you have a curve, the integration of which you approximate by a series of rectangles. At n=5 you get one answer, at n=10 you get a better answer. As n approaches infinity the approximation converges with the actual area under the curve.

Your data even shows this trend. It looks like the two answers would converge if you continued refinement.

[jseamon1]

Oh, that makes sense. Thank you so much! I appreciate the insight!