I am doing research on Ranque Hilsch vortex tube which is a 3D, turbulent and compressible problem. I am using K-Epsilon (2 equation) model(before that I tried RSM). And my solver is coupled implicit and steady state. Currently, in the beginning the residuals are decreasing (upto 600 iterations) but then they stay at a certain constant level. I want complete convergence with accuracies atleast 10-06. Any hint. Thanks in advance.

Are you runing in Double Precision or in Single?

Yes, I am running double precision.

Do a search of the forum for convergence and residuals. The actual value on the residuals plots don't mean much because they're normalized values. If your initial solution is reasonably close to your final solution, then your residuals will not drop much (people seem to forget this when using the FMG initializer... the initial solution is pretty close to the final, so their residuals don't drop much). You should be judging convergence by the shape of the residual plots (leveling off is a better judge of convergence than reaching a certain level), but precision should be judged using monitors (either force, local pressure, average pressures, etc... whatever you plan on reporting as a result, you should be monitoring).

Hope this helps, and good luck, Jason

[atulurade]

sir, please tell me about the boundary conditions setting

Originally Posted by Sachin U. Nimbalkar
;137648

I am doing research on Ranque Hilsch vortex tube which is a 3D, turbulent and compressible problem. I am using K-Epsilon (2 equation) model(before that I tried RSM). And my solver is coupled implicit and steady state. Currently, in the beginning the residuals are decreasing (upto 600 iterations) but then they stay at a certain constant level. I want complete convergence with accuracies atleast 10-06. Any hint. Thanks in advance.

[jjfm20]

Hello,

Sorry for comment this old post, currently I'm having exactly the same problem, I'm using RSM, Ideal Gas and Kinetic Theory for determining the thermodynamics properties, also I'm using a tet mesh because I wasn't capable of meshing that with hex in an acceptable quality...

I tried all kind of residuals, about 1million elements and still I can't get any good convergency...

If you solved I would be very grateful if you tell me what should I do.

Thanks