Under-relaxation factors

L3munoz · May 21, 2009, 15:06

Hi!

I was running a simulation and the convergence was quite difficult, having lots of picks and changes. So I decided to decrease the under-relaxation factors, and now the convergence is going slower but much more smooth.

Since now there are less changes and it seems that it has almost convergenced (CD values didnt change a lot in the last iterations), should I consider that convergence is already achieved or this is because of reducing these factors?? Which is the effect of reducing them on the convergence, just smoothing it or making it to take more time because changes are smaller? Or, by the opposite, it helps to achieve convergence faster?

Thanks in advances,

Jorge.

Dmitry · May 22, 2009, 06:55

Hi!
As I know, increasing under-relaxation factors, gives faster convergence, but your solution may become unstable, even it can diverge.
Decreasing factors gives slower convergence, but better stability.

L3munoz · May 23, 2009, 04:03

Hi Dmitry!

Thanks! But my question wasnt answered totally...I already saw that decreasing under-relaxation factors, convergence is slower. However my criteria to determine convergence is CD value. And this didnt change too much in a big period of iterations. Should I consider it is already converged or by the opposite, this is just because of decreasing these factors and making changes to appear slower??

Ciao!

Giuki · May 23, 2009, 10:29

Hey Jorge,

I haven't been using fluent for aerodynamic reasons, but mainly for combustion simulation, however I'm working in the same room as other students and as I've understood: When the value of Cd is not changing anymore (e.g. levelled out (straight line)) you may consider the simulation converged. To be sure I'll ask them on monday, but I'm pretty sure!

Regards!

Dmitry · May 25, 2009, 04:53

If you simulation is canal flow, you can look at mass flow at inlet and outlet. If they are not close to each other, your solution is not converged, even your residuals are low.

L3munoz · May 25, 2009, 06:32

Giuki: Hi!! Yes, my idea was that if CD value is almost constant, convergence is achieved...however it is changing a lot. Ok, not a lot, but it is observed some picks and the value is nor among a range, but still decreasing. The point is that after 7000 iterations, it didnt converged... I'm working with 6.10e6 cells, but... If you could ask something to ur collegues, I'd be very pleased of get some information... Thanks in advance!

Dmitry: Hi!! No, my simulation is an openfield one. I'm trying to get the cd over a train model, so I guess the way is just by looking at the CD value...But as I was saying before, it is observed some picks and I dont know if 7000 iterations are already too much for this amount of cells or not... I dont know if to stop the simulation or still wait for it....

sivam · May 27, 2009, 02:31

Quote:

Originally Posted by Dmitry

Hi!
As I know, increasing under-relaxation factors, gives faster convergence, but your solution may become unstable, even it can diverge.
Decreasing factors gives slower convergence, but better stability.

U have posted a valid point, but in our case we decreased under relaxation factor ,which make our solution a converge fast and get a required result,My question is will this provides a good and accurate result, can we relay on the results or else we have to do any processing on it later.Thanks in advance.

anupamjain76 · May 29, 2009, 05:04

Hi,

The convergence should be judged on the basis of following things:
(1) Overall mass and energy conservation should be achieved.
(2) Some important flow parameters(for eg. velocity, temperature, pressure, Cd, etc) should be monitored at locations of interest and they should not change with iterations.

If the above two criteria are met then your solution is converged.

Anupam

May 21, 2009, 15:06	Under-relaxation factors	#1
L3munoz Member Jorge Muñoz Join Date: Apr 2009 Posts: 56 Rep Power: 17	Hi! I was running a simulation and the convergence was quite difficult, having lots of picks and changes. So I decided to decrease the under-relaxation factors, and now the convergence is going slower but much more smooth. Since now there are less changes and it seems that it has almost convergenced (CD values didnt change a lot in the last iterations), should I consider that convergence is already achieved or this is because of reducing these factors?? Which is the effect of reducing them on the convergence, just smoothing it or making it to take more time because changes are smaller? Or, by the opposite, it helps to achieve convergence faster? Thanks in advances, Jorge.

Thread Tools	Search this Thread
Show Printable Version Email this Page	Search this Thread: Advanced Search
Display Modes
Linear Mode Switch to Hybrid Mode Switch to Threaded Mode

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Relaxation Factors for Transient solvers	philippose	OpenFOAM Running, Solving & CFD	19	March 20, 2014 04:39
relaxation factors and time accuracy	Mike	Main CFD Forum	7	May 21, 2005 12:41
Relaxation Factors	Tim	Phoenics	3	June 30, 2004 02:03
relaxation factors	zhujianguo	Phoenics	1	August 5, 2003 06:17
relaxation factors adjust	zhujianguo	Phoenics	1	July 15, 2003 11:11

May 22, 2009, 06:55		#2
Dmitry New Member Dmitry Join Date: Mar 2009 Posts: 7 Rep Power: 17	Hi! As I know, increasing under-relaxation factors, gives faster convergence, but your solution may become unstable, even it can diverge. Decreasing factors gives slower convergence, but better stability.

May 23, 2009, 04:03		#3
L3munoz Member Jorge Muñoz Join Date: Apr 2009 Posts: 56 Rep Power: 17	Hi Dmitry! Thanks! But my question wasnt answered totally...I already saw that decreasing under-relaxation factors, convergence is slower. However my criteria to determine convergence is CD value. And this didnt change too much in a big period of iterations. Should I consider it is already converged or by the opposite, this is just because of decreasing these factors and making changes to appear slower?? Ciao!

May 23, 2009, 10:29		#4
Giuki New Member Giuki Cael Join Date: May 2009 Location: Belgium/The Netherlands Posts: 13 Rep Power: 16	Hey Jorge, I haven't been using fluent for aerodynamic reasons, but mainly for combustion simulation, however I'm working in the same room as other students and as I've understood: When the value of Cd is not changing anymore (e.g. levelled out (straight line)) you may consider the simulation converged. To be sure I'll ask them on monday, but I'm pretty sure! Regards!

May 25, 2009, 04:53		#5
Dmitry New Member Dmitry Join Date: Mar 2009 Posts: 7 Rep Power: 17	If you simulation is canal flow, you can look at mass flow at inlet and outlet. If they are not close to each other, your solution is not converged, even your residuals are low.

May 25, 2009, 06:32		#6
L3munoz Member Jorge Muñoz Join Date: Apr 2009 Posts: 56 Rep Power: 17	Giuki: Hi!! Yes, my idea was that if CD value is almost constant, convergence is achieved...however it is changing a lot. Ok, not a lot, but it is observed some picks and the value is nor among a range, but still decreasing. The point is that after 7000 iterations, it didnt converged... I'm working with 6.10e6 cells, but... If you could ask something to ur collegues, I'd be very pleased of get some information... Thanks in advance! Dmitry: Hi!! No, my simulation is an openfield one. I'm trying to get the cd over a train model, so I guess the way is just by looking at the CD value...But as I was saying before, it is observed some picks and I dont know if 7000 iterations are already too much for this amount of cells or not... I dont know if to stop the simulation or still wait for it....

May 29, 2009, 05:04		#8
anupamjain76 New Member Anupam Jain Join Date: May 2009 Posts: 5 Rep Power: 16	Hi, The convergence should be judged on the basis of following things: (1) Overall mass and energy conservation should be achieved. (2) Some important flow parameters(for eg. velocity, temperature, pressure, Cd, etc) should be monitored at locations of interest and they should not change with iterations. If the above two criteria are met then your solution is converged. Anupam