Choosing convergence criteria on monitor point
I am monitoring torque in a rotating component. The torque value I am expecting is between 1000-2000 N*m, depending on rotational velocity. I would need to find the value within 25 N*m. What should I use to determine when this happens, as it seems there will always be a slight change if i let it go on forever. Would something like a change of less than .01% over 20 iterations likely be sufficient for what I'm looking for?
You are looking for quiet accurate results (2.5 on torque) so you are going to have to be very careful with accuracy to achieve this. I think you are going to have to be a bit more clever than that to achieve your target accuracy reliably.
I would have a look at the key reference text on CFD accuracy "Computational Fluid Dynamics" by Roache. Things like the link on this FAQ are a good summary http://www.cfd-online.com/Wiki/Ansys...publishable.3F
A thought on your specific case of judging convegernce: If you let your case converge for a while you may find that the torque returned is following a relationship like error=1/x^n or error=e^-nx or something. If you can pick what path it is taking for convergence you can get a far more accurate torque number than you simple tolerance approach. Note that this is a similar approach as Richardson extrapolation is to mesh refinement studies.
|All times are GMT -4. The time now is 12:36.|