Discussion Groups

## Simulation CFD

Contributor
Posts: 15
Registered: ‎10-16-2012

# CONVERGENCE CRITERIA

316 Views, 2 Replies
12-04-2012 02:43 AM

Hi all, i'm a new user of CFD 360. I've been performing simulation with CFX before. I've learned that CFD 360 uses a pseudo transient approach for solve steady state problems. So my question is: to get the same convergence criteria of CFX what should i consider, the RES IN or the RES OUT? My idea is this: since it's a pseudo transient process i should consider the RES OUT because it's the final RES of the pseudo transient iterations, what I was considering before when i solve transient anakysis in CFX.Is this correct??

Hope i've been sufficiently clear, thanks in advance for any kind response.

*Expert Elite*
Posts: 406
Registered: ‎10-02-2012

# Re: CONVERGENCE CRITERIA

12-04-2012 06:37 AM in reply to: neri69

neri69,

SimCFD uses pseudo transient solver only when you enable "Intelligent Control" in Sovle dialogue box, and alters the timesteps as solution marches ahead. But interpretation of residual is tricky.

From what I know, the CFX (and Fluent) uses the normalised residuals. Such that at the end of all iterations, it divides the actual residual (difference between RHS and LHS of equation for given variable) and normalises it by LHS etc. While in SimCFD, Resid In and Resid Out are residuals before and after convergence. Essentially, even if you take Resid Out, which is the difference between LHS and RHS for a variable for the last iteration, you still don't have a normalised residual (comparable to CFX).

What SimCFD recommends as a convergence criteria is that the ratio (Resid out/Resid In) should be less than 1e-6, and the last 20% of the iterations should show no more than 5% change in the parameters (like pressure, velocity on monitor point etc.)

Of course, more experienced users may have more to add here.

Cheers

OJ

Contributor
Posts: 15
Registered: ‎10-16-2012

# Re: CONVERGENCE CRITERIA

12-04-2012 06:56 AM in reply to: OmkarJ

Thank you very much Omkar, you've been really clear, so what i should consider is the ratio between RES out and RES in and any kind of oscillation in dependent variables values in the last iterations!

Thank you!!