convergence and termination criteria in acusolve

Unknown
Unknown
edited July 2023 in Community Q&A

Hi guys,

I have some questions regarding the termination criteria and convergence level in acusolve.

1. I know the plots (residual ratio and solution ratio) are, by default, logarithmic in scale. I changed them back to linear. The linear plot is almost horizontal. Does this means convergence?

2. As far as I know, the termination criteria in acusolve are:

                             for steady:- max time steps or convergence tolerance

                             for transient: max time steps, final time or convergence tolerance.

I have specified 0.001 as convergence tolerance. But the simulation runs continuously till time = 5 s. It does not stop until 5 s. How to know that velocity and pressure has met specified tolerance level??? I can write them to a file using convert tool. Also, please explain how to determine convergence in CFD.

image

I wrote the pressure at inlet to file:

image

The following is the surface integrated heat flux:

image

They are correct upto like 6 significant digits (which is what I want). Am I judging convergence correctly in CFD?

I look for your meaningful comments.

Thanks.

Tagged:

Answers

  • acupro
    acupro
    Altair Employee
    edited July 2023

    You may want to look at this FAQ in the knowledge base:
    https://community.altair.com/community?id=kb_article_view&sysparm_article=KB0122435

    That should cover most of your questions.

    When running transient, the 'convergence' is on a per-time-step basis.  We indicate the maximum time steps and/or maximum time - whichever comes first.  Then each time step is judged for convergence based on residual-ratio and solution-ratio values.

    Based on those plots, I would guess there's not much change in solution after time = 0.5 sec or so.  Convergence based on convergence-tolerance is reached very well from that point on.

    As in that post mentioned above - residual/solution ratio is only one method to judge the simulation.  Best also to see how results of interest are behaving.

  • Unknown
    Unknown
    edited July 2023
    acupro_21778 said:

    You may want to look at this FAQ in the knowledge base:
    https://community.altair.com/community?id=kb_article_view&sysparm_article=KB0122435

    That should cover most of your questions.

    When running transient, the 'convergence' is on a per-time-step basis.  We indicate the maximum time steps and/or maximum time - whichever comes first.  Then each time step is judged for convergence based on residual-ratio and solution-ratio values.

    Based on those plots, I would guess there's not much change in solution after time = 0.5 sec or so.  Convergence based on convergence-tolerance is reached very well from that point on.

    As in that post mentioned above - residual/solution ratio is only one method to judge the simulation.  Best also to see how results of interest are behaving.

    "I would guess there's not much change in solution after time = 0.5 sec or so. " How to determine this? Did you see the 10^-3 on y-axis of the residual ratio?

  • acupro
    acupro
    Altair Employee
    edited July 2023
    Prabin Pradhananga_20428 said:

    "I would guess there's not much change in solution after time = 0.5 sec or so. " How to determine this? Did you see the 10^-3 on y-axis of the residual ratio?

    The residual ratio and solution ratio values are 1.e-8 and lower.  Generally if the solution were still changing from time step to time step the values would be in the 1.e-3 range and lower.

    It's interesting that convergence (based on convergence tolerance, etc) is not really reached until time 0.5 s or so.  You may want to focus a bit in those earlier times to make sure the model results are as expected.  Maybe increase the max-stagger-iterations a bit?

Categories