CFD Online Discussion Forums

CFD Online Discussion Forums (
-   FLUENT (
-   -   Need desperate help in convergence (

Puck August 1, 2002 22:19

Need desperate help in convergence
I have a new problem now... I find that though my geometry is symmetrical(but im not using symmetrical planes) my solution looked weird when it converged.. there was no symmetry in the solution for heat transfer.. Therefore, i was asked to decrease the convergence criteria halfway through my iterations..i decreased all my parameters by 0.001 from the default values.. Now i find that those parameters which i decreased e.g z velocity is decreasing more than the criteria i input..The convergence criteria for all the parameters is around 10 to the power -5...meaning 0.000001.. But now the iterations show 0.00000001 and still havent converge.. I only set my energy as 10 to power -8..which is still in that range.. I stopped the iterations to see if my heat profile has symmetry; it has...So wht should i do? Assume the solution has converged? Plz help...

Erwin August 2, 2002 05:39

Re: Need desperate help in convergence
Residuals are only one way of checking convergence, but they are not really what it is all about. You cannot specify a lower limit that residuals MUST adhere to, makes no sense. I have had problems with radiation in the past where the residuals did not change for a thousand iterations but the radiation fluxes kept on changing...

So, look at your actual solution, if it looks okay (like have symmetry where you expect it) and if it doesn't change significantly during a number of iterations, then your solution is converged. Period.

Puck August 2, 2002 09:58

Re: Need desperate help in convergence
Hi... thnx for replying... So u mean to say that, as long as i see a symmetrical profile, i can stop the iterations without having the software saying'solution converged?

Erwin August 3, 2002 13:05

Re: Need desperate help in convergence
Yes, the convergence criteria are artifical. Use your brain, look at your solution, decide for yourself, don't wait for the software to come up with something intelligent.

Puck August 4, 2002 01:22

Re: Need desperate help in convergence
okie thnx.. will do tht in future.

Alamgir August 11, 2002 21:23

Re: Need desperate help in convergence
10^(-5)=0.00001, not 0.000001. Erwin is right, u can stop after getting your desire profile don't bother with the residuals or saying "equation is converged".


All times are GMT -4. The time now is 02:58.