CFD Online Discussion Forums

CFD Online Discussion Forums (http://www.cfd-online.com/Forums/)
-   FLUENT (http://www.cfd-online.com/Forums/fluent/)
-   -   help in relaxation factor (http://www.cfd-online.com/Forums/fluent/37523-help-relaxation-factor.html)

SONALI August 9, 2005 03:07

help in relaxation factor
 
i want to know whether the solution is said to be converged when we lower the relaxation factor than the default values or for getting the accurate solution it is nessary to use the default values of relaxation factors.thanks a lot.

zxaar August 9, 2005 03:18

Re: help in relaxation factor
 
relaxation factors only alter the route to converged solution, and should not have any effect on finally converged value.

that means a converged solution obtained from factor of 0.0001 and 0.9 will be same. but with urf of 0.0001 solver will take lot of time to converge.

usually the defaults values are good except for where fluent makes them 1, they shall be changed to 0.9 or lower, to get stable solution.

since higher urf sometimes may make solution unstable.


Swarup August 9, 2005 11:10

Re: help in relaxation factor
 
I read in flow3d.com that too much under-relaxation is not right for implicit linearization. although I have not faced such a situation, I would like to know if this is really that serious and would be grateful for any advice in this regard.

Swarup

zxaar August 9, 2005 20:11

Re: help in relaxation factor
 
yes and no , there are two ways one can implement the under relaxation factors, usually :

ap = ap / urf

su = su + (1-urf) * ap * phi

and calculate new phi

and another way is :

calculate new phi and

phi = phi_old + urf * (phi_new - phi_old)

in the first way, we can increase the diagonal dominance and hence we can increase the convergence. so in this way if you increase the urf too much, ap might be unnecesaryly increased and may goof up the solution altogether.

but fluent uses the second way and selecting urf do not interfere with solution matrix, so you can use very small urfs.


Mahesh August 11, 2005 06:29

Re: help in relaxation factor
 
I dont think urf will have any effect on the final converged solution, except the speed of the analysis ( Assuming that solution is not diverging)

zxaar August 11, 2005 20:27

Re: help in relaxation factor
 
well with very small urf the problem is: the assumption of not diverging( if you are using the first method).

in a small cube 1 x 1 x 1 with say 1m/sec inlet, the typical ap value is around 0.5, now in first manner

if urf is 0.00001,

ap_new = ap / urf = 5 X 10^4, and

su_new = su + (1- 0.00001) * ap_old * phi

note that su does not change so much, by seeing the change in ap its very much clear that you have already screwed the original problem, now if the solution converges you are very lucky.

so with so small urf assuming that you will get converged solution this way is somthing i can not agree with.

yes the second way, when we apply the urf it does not fiddle with matrix structure and the solution convregnce may become slow, but it will not diverge because of urf.

and yes convregd solution is convregd solution it shall not depend upon the route it takes to reach it.


All times are GMT -4. The time now is 14:26.