Single Phase simulation of an ejector
 

Hi -

I am trying to replicate a technical paper which validated experimental results with a single phase R245fa ejector. I am having a lot of issues getting the solution to converge. I am using the NIST real gas module in Fluent for R245fas simulations. Summary of the model:

2d, planar, density, k-w with SST. The motive nozzle inlet is single phase vapor with a pressure of 4 atm and temperature of 363.15 K. The suction inlet (outlet of the evaporator in a vapor compression cycle) of 0.6 atm and 303.15. The pressure outlet condition of the ejector is 0.1 atm (303.15K).

I am using the implicit Roe scheme (least squares method) second order terms for flow and turbulence. I have lowered the Courant number to 1 (since solution diverges if I go higher than 5) and the "under-relaxation" factors are 0.5.

Currently I don't have mass flow rate convergence at the ejector outlet even after 3700 iterations. Any help is greatly appreciated!!!

3700 iterations is nothing (not very many).

Complex models are generally prone to divergence and even when they don't diverge, can converge slowly.

In general it's very hard to get a real gas simulation to converge because of non-linearities. You need to have a very good initial guess. You also need to tune the urf's a lot.

Are you using a targeted mass flow rate? Instead, use a fixed pressure inlet and fixed pressure outlet. This helps a lot.

I hope you've already tried starting with a slightly simpler simulation (constant properties, or only temperature dependent properties) before going to the full blown NIST real gas.

Using a predefined lookup table also speeds up the computation tremendously if you need to shave some compute hours.

In general it's very hard to get a real gas simulation to converge because of non-linearities. You need to have a very good initial guess. You also need to tune the urf's a lot.

That's a good point. I had that difficulty initially. What I did was to try and solve the problem in 1-D, and used the results from that in my 2-D model. Convergence hasn't improved though (though to your first point, 3700 is not high. Currently, it is at 9000 iterations, and continuity, momentum and energy are still at 10^-1. MFR is oscillating between reversed flow from the outlets and back.

Are you using a targeted mass flow rate? Instead, use a fixed pressure inlet and fixed pressure outlet. This helps a lot.
I am employing a fixed pressure inlet and outlet conditions. I tried the targeted MFR earlier to see if convergence was better, but made it worse.

I hope you've already tried starting with a slightly simpler simulation (constant properties, or only temperature dependent properties) before going to the full blown NIST real gas.
Actually, I tried this with air, with no solution convergence issues. This is great advice, I tried to short myself and probably wasted time. I will try an ideal gas model right away, and post back here.

Using a predefined lookup table also speeds up the computation tremendously if you need to shave some compute hours.
I do use a pre-defined lookup table as advised in the real-gas information via ANSYS.

Thanks LuckyTran!

Help
 

Dear Lexicon

Does it work with you?
if yes kindly advice with any useful tips since i'm working on 2d-axisymmetry modeling using NIST real gas for R134a and i'm not getting any convergence
also i'm getting REFPROP_error (203) from function: tprho (density)
[TPRHO error 203] vapor iteration has not converged

if anyone can help i will be thankful

Best Regards
Moh'd