Use the following best practices to improve the accuracy of steady coupled analyses.
- Define an appropriate number of coupling steps/iterations.
Best practices for choosing the number of coupling steps and/or iterations vary according to whether the analysis is iteration-based or step-based. (For more information, see Steady Analysis Type.)
Iteration-based steady analyses:
An iteration-based steady analysis allows you to define the minimum and maximum number of iterations. Restart points can be created at the end of each coupling iteration. Ramping and under-relaxation can be applied across iterations.
Ensure that you define enough coupling iterations to allow the analysis to converge. For more information, see Assess the convergence status of the analysis.
Step-based steady analyses:
A step-based steady analysis can have one or more coupling steps, each with multiple steps permitted. Restart points can be created only at the end of each coupling step.
Set the number of coupling steps and coupling iterations according to how you need to balance requirements for restarting the analysis, minimizing file-storage space, and/or using System Coupling's ramping and under-relaxation algorithms.
Ramping capabilities and setup recommendations vary according to the number of coupling steps, as follows:
One coupling step:
Because restart data are generated only at the end of a completed coupling step, a single-step analysis can only be restarted at the end of the solution. As such, it is not possible to restart from an intermediate result or an abnormally terminated run.
Minimizes storage space at the expense of restart capabilities.
Allows ramping and under-relaxation across coupling iterations within the single coupling step.
Multiple coupling steps:
Restart data are still generated only at the end of coupling steps, but you can specify which steps for which restart data will be generated. As such, it is possible to restart from an intermediate result or from an abnormally terminated run, which can be useful for complex analyses that are at risk of abnormal terminations.
Allows multiple restarts at the expense of storage space.
Allows ramping and under-relaxation across coupling iterations within a coupling step, but not across coupling steps. When a step is completed, the full data transfer value is transferred to the next step.
- Assess the convergence status of the analysis.
Identify and address any issues that may be affecting the convergence of the analysis.
Slow convergence:
If convergence is not realized within tens of coupling iterations (that is, up to 100), then consider introducing stabilization methods such as ramping and under-relaxation.
Participant-specific solution stabilization methods, such as those offered by Fluent, may also be employed. For details, refer to the corresponding participant product documentation.
Strong non-linearities:
If there are strong non-linearities within the coupled solution, then more coupling iterations may be required, even with the use of stabilization. You can increase the maximum number of coupling iterations per step. For step-based steady analyses, you can also try introducing additional coupling steps.
Premature convergence:
In some cases, convergence is realized prematurely (for example, within one iteration) because the non-linearities in the coupled solution have not yet been activated. If this is suspected, then set a minimum number of coupling iterations per step to trigger the non-linearities and force adequate resolution.