relaxation factor
Hi experts,
I am having trouble with the covergence in the acusolve simulation. I read about the relaxation factor and tried with few values and here are the results. I am seeking for your comments about which result is good in terms of stability. Also, I am looking some suggestions about convergence just by looking at these plots. (All the simulations are equivalent to each other except the relaxation factor).
The first image (rf = 0.1) seems to be stable but the solution ratio is not changing at all, which means it may not have converged.
I want to avoid wiggles that I see in the second image (rf = 0.6).
May be the third image is good. I may need to decrease the convergence tolerance (rf = 0.8). Also note that this simulation has converged at time step ~66.
Thanks.
Best Answer
-
Based on those plots with 0.6 relaxation factor, I would guess the flow really is transient in nature, and that a steady solution is likely not physical.
I would recommend not going above 0.7 with relaxation factor.
2
Altair EmployeeAnswers
-
Based on those plots with 0.6 relaxation factor, I would guess the flow really is transient in nature, and that a steady solution is likely not physical.
I would recommend not going above 0.7 with relaxation factor.
2 -
acupro_21778 said:
Based on those plots with 0.6 relaxation factor, I would guess the flow really is transient in nature, and that a steady solution is likely not physical.
I would recommend not going above 0.7 with relaxation factor.
I changed the problem from steady to transcient without making any other changes. The plot looks like the following. I seek some comments about stability and convegence. Do I need to run for more time? The convergence tolerance is 1e-4.
0 -
Prabin Pradhananga_20428 said:
I changed the problem from steady to transcient without making any other changes. The plot looks like the following. I seek some comments about stability and convegence. Do I need to run for more time? The convergence tolerance is 1e-4.
Those convergence plots look better. You need to pay attention to your solution(s) of interest, too. How are those changing/behaving? You need to decide if the overall simulation makes sense.
1 -
acupro_21778 said:
Those convergence plots look better. You need to pay attention to your solution(s) of interest, too. How are those changing/behaving? You need to decide if the overall simulation makes sense.
Thanks for your comments. My solution of interest (pressure drop and surface integrated heat flux) are very stable than that of steady simulation. The velocity field is now making more sense than before.
0 -
Prabin Pradhananga_20428 said:
Thanks for your comments. My solution of interest (pressure drop and surface integrated heat flux) are very stable than that of steady simulation. The velocity field is now making more sense than before.
Hi acupro,
I had a follow up question regarding the steady and transient simulation.
The problem we are solving is steady state in nature. We found, in the literature, most people assume steady state and perform the CFD simulation. They get good result with this assumption. However, I tried the same with steady and I have convergence issues. But when I shifted to transient, I have convincing results.
I am thinking why this problem is not working in steady state. Could there be any justification for this?
0 -
If you increase the relaxation factor, the solution tends to be more unstable right but its other way round in your case. Just waned to clarify if you are increasing Steady update factor?
1 -
Jagan_21383 said:
If you increase the relaxation factor, the solution tends to be more unstable right but its other way round in your case. Just waned to clarify if you are increasing Steady update factor?
Hi Jagan,
I have used the default values for steady update factor of 0.6. Should I try changing the steady update factor?
0 -
Prabin Pradhananga_20428 said:
Hi Jagan,
I have used the default values for steady update factor of 0.6. Should I try changing the steady update factor?
Steady update factor is nothing but 1- relaxation factor, if its 0.6 your relaxation factor is 0.4.
1 -
Jagan_21383 said:
Steady update factor is nothing but 1- relaxation factor, if its 0.6 your relaxation factor is 0.4.
Thank you. In the steady state simulation, I have steady update factor of 0.6. When I open the input file for steady simulation, there's no steady update factor but there's relaxation factor of 0.4.
0 -
Prabin Pradhananga_20428 said:
Hi acupro,
I had a follow up question regarding the steady and transient simulation.
The problem we are solving is steady state in nature. We found, in the literature, most people assume steady state and perform the CFD simulation. They get good result with this assumption. However, I tried the same with steady and I have convergence issues. But when I shifted to transient, I have convincing results.
I am thinking why this problem is not working in steady state. Could there be any justification for this?
If you go back to steady - with relaxation factor 0.3 or 0.4 (steady update factor 0.7 or 0.6 in HWCFD GUI) - maybe request more frequent nodal output so you can see if there are locations where the solution is changing. Those earlier plots with relaxation factor 0.6 showed the velocity was still changing quite a bit - solution ratio on the order of 0.1.
Again - try to determine where the solution is changing by looking at the results in Post Processing.
1 -
Prabin Pradhananga_20428 said:
Thank you. In the steady state simulation, I have steady update factor of 0.6. When I open the input file for steady simulation, there's no steady update factor but there's relaxation factor of 0.4.
You are right. In input file there will be only relaxation factor (based on steady update factor you define in Physics tab).
0