trying to simulate flow through a tube with a 50 micron orifice at the center. the tube diameter is 0.19 in (~5 mm). solving full 3d. the fluid is ethanol and the mass flow rate is 3.333e-4 kg/s. the exit pressure is 100 psig.

i'm getting a 480 psig drop in pressure across the orifice and a peak velocity of 63 m/s. From continuity analysis (Q/A) we get an average velocity of ~215 m/s. so, there's a mismatch

the continuity residuals are down to 1e-10 (using the unsteady segregated solver) and the momentum residuals are down to 1e-7. i'm using the k-e model and the k residuals are down below 1e-6, the epsilon residuals are around 1e-4. have turned on the energy equation and the residuals are of the order of 1e-6.

i'm unable to figure out what is wrong as the centerline velocity inside the orifice should be greater than the average velocity which was calculated using the Q/A continuity analysis.

possible causes:

1. the orifice diameter being small (0.05 mm, area ratio [5mm dia to 0.05 mm dia] = 10,000), the flow is not being captured properly. the continuity residuals are below 1e-9 everywhere inside the orifice.

2. flow in the massive recirculation zones upstream and downstream of the orifice is not being captured properly.

3. the velocity contours inside the orifice are not uniform. the length of the orifice is 0.5 mm and i've taken 10d upstream and 10d downstream length (d is the tube diameter)

Any suggestions as to what is wrong ? i'm pretty sure that the boundary conditions are all right. i've checked the continuity across the geometry and it has come out to around 1e-9. should the momentum residuals go down to even lower values. is it that the jet coming out of the orifice into the tube hasn't formed yet ?

Why not run steady state at first and see if the results match your continuity analysis?

If the steady state results are more sensible, do a few grid independence studies (run with finer grids).

Regards,

Chinor

I did run a steady case. Residuals for continuity were two orders higher and so were the momentum and other residuals.

I have a grid with 150,000 node points. Am using an SGI Origin 2000, four 195 MHz processors. Refining the grid won't be possible without sacrificing the computational time requirement. I've noted that even though I have very fine grid near the orifice (both upstream and downstream), the transition of the cells is abrupt.

We need an answer fast for a senior design project class who are waiting to order equipment based on my calculations.

OK, I am going to step out on limb and really show some of my inexperience; however, I should at least simulate some discussion that will speed your senior design.

1. I will agree your CFD numbers are a bit-off; however, is the Q/A analysis the best estimate for this type of problem? What about the viscous losses in the sudden contraction?

2. Based on the extreme differences in velocity between sections, would the effects of turbulences not be limited allowing you to assume a laminar flow and speed up your simulations?

>1. I will agree your CFD numbers are a bit-off; however, is the Q/A analysis the best estimate for this type of problem?
:What about the viscous losses in the sudden contraction?

Q/A analysis holds. Viscous losses in the sudden contraction region will not influence the average velocity inside the orifice. There will be an associated pressure loss though (which was found to be negligible because of the very low velocities in the recirculation region).

>2. Based on the extreme differences in velocity between sections, would the effects of turbulences not be limited

>allowing you to assume a laminar flow and speed up your simulations?

The flow inside the orifice is mildly turbulent and yes, the simulation "may" (with lots of doubts) be done with a laminar solver.

I have been able to correct the error which resulted in the disparity with the Q/A analysis. The problem was the sudden change in the tetrahedral cell sizes from the tube to the orifice. I had tried to keep the grid near the orifice entrance fine, but the computational requirement restriction had bogged me down.

Thanks for the suggestions though. I've been able to simulate the flow quite accurately now.