Learning Forum

[164170007]

Hello Everyone,

I would like to know the influence of under relaxation factors.

I did two same simulations with different relaxation factors. (Laminar, steady state case)

a) In first simulation I am using COUPLED scheme and for initial 100 iterations I used 0.25 for species and energy and other is by default and after 100 iterations when reversed flow is over I changed from 0.25 to 0.99 and it converges very fast (within 1 hr) with proper downward pattern (residuals 1e-9).

b) In second simulation I have choosen 0.95 relaxation factor for same case and solver from first iteration to till now, but its not converged yet for same residuals (1e-9) (more than 20 hr simulation).

My question is that 1) Is it a right practice to change relaxation factor in middle of the simulation?

2) Is it going to affect the final result?

Thanks in advance.

Regards,

Vinay

 

[Essence]

 

 

Hello,

Under-relaxation Factors (URFs) need to be changed only if necessary or the solution is unstable/diverging initially. The default URF values provided in the Ansys Fluent are good enough for most of the simulations. But while increasing them, the solutions may get faster but it might decrease the stability, which sometimes might lead to divergence. That’s why care must be taken while changing them. Of course, it is possible to change the URFs in the middle of the calculations. But if the URFs are set quite low, the stability will be high but it might take more number of iterations for the solution to converge. The end result will not be affected given, the proper setting of the case is done.

 

Please refer more here on the URFs: 32.20. Checking Your Case Setup (ansys.com)

14.2.4. Stability and Convergence (ansys.com)

If you are unable to access the link, follow this Forum discussion https://forum.ansys.com/forums/topic/using-help-with-links/#latest

 

 

[164170007]

Hello Essence,

In my simulation initially divergence occurs due to the reversed flow and when I reduced the relaxation factors the problem of reversed flow has gone and then I changed URF (by default) and residuals pattern are also stable and converged faster.

[Essence]

Did it solve your problem?

[164170007]

Yes, results are similar for both case and the second test took higher computational time.

[Essence]

Please note that, just by having low scaled residuals does not mean the solution is converged. You should also try the net mass flow rate and energy balance (if needed). Additionally, observe the residuals for the flow properties (eg: pressure) if they have converged or not. You might need to set the residuals for the flow properties manually. 

Your solution might get converged even if the residuals are 1e-3. But, the mentioned conditions for the convergence needs to be checked.

[164170007]

I kept residuals upto 1e-9 for better accuracy and its converging successfully. Thank you for your help.

Regards,

Vinay