Automatic Convergence Assessment determines when a solution is converged--when the solution stops changing--and automatically halts the calculation. It examines small and large frequency changes throughout the solution field, and evaluates the local and global fluctuations of each degree of freedom.
Automatic Convergence Assessment is automatically enabled for the same types of analyses as Intelligent Solution Control. Automatic Convergence Assessment is enabled or disabled by clicking the Advanced button on the Solution controls dialog, and checking or unchecking the Automatic Convergence Assessment box.
Automatic Convergence Assessment removes the guess-work of knowing when a solution is completed. Four different parameters are evaluated, and the threshold criteria levels can be changed with the slider bar. By default, the criteria are set to be moderate--between Loose and Tight. This will provide reasonable convergence for a wide variety of analysis types. “Reasonable” means that the convergence criteria are rigorous, but not exhaustively so. They will consider a 1% variation in the Summary trends to be converged. This is appropriate for most analyses with exceptions listed below.
Change the slider setting to Loose for a “preliminary” analysis in which extremely high accuracy is not the goal. Such analyses are very useful for identifying trends in a design. Convergence will typically occur with fewer iterations, but the results may not be as accurate. Change the slider to Tight to invoke more rigorous convergence criteria. This is useful for a final analysis in which a high level of convergence and accuracy is necessary.
It has been observed that in some analyses in which aerodynamic- or hydrodynamic-induced forces are of interest, the solution may be considered converged and stopped by Automatic Convergence Assessment before the forces have actually stopped changing. The forces in such analyses (such as aerodynamic flows over thin bodies) often require many hundreds of iterations to reach fully converged force values, and may require additional iterations beyond where Automatic Convergence Assessment will stop the calculation. In such cases, it is recommended to disable Automatic Convergence Assessment and run additional iterations. Monitor the forces manually to ensure that they have stopped changing.
Additionally, flows that rely only on shear drag for their pressure drop, such as flow through a pipe, tend to require more iterations to converge. In such analyses, the default slider setting may halt the calculation prematurely. For this reason, it is recommended to change the setting to Tight for pipe flow analyses that do not contain any form-drag.
Reliance on Automatic Convergence Assessment is not recommended for transient analyses that will not reach a steady-state solution such as Rotating, Motion, or vortex shedding analyses. By their nature, none of these types of analyses will ever typically reach a numerically converged state that satisfies Automatic Convergence Assessment. For this reason, it is recommended that the stopping criteria be evaluated manually based on the desired time span of the analysis or other physical objective.
Auto-Stop Criteria
Several convergence criteria to indicate a converged solution and stop the analysis. These criteria are:
For this criterion, the slopes of the convergence data of quantities on the Convergence Monitor are evaluated from one iteration or time step to the next. The minimum, maximum and mean value of all of the dependent variables is examined. When the maximum instantaneous slope in all of this data is below the set level, the solution is stopped.
The slope of the convergence data over several iterations or time steps is evaluated. The minimum, maximum and mean values of all of the dependent variables are considered.
The derivative of the maximum time averaged convergence slope is evaluated. This derivative is a measure of the concavity or whether the curve is flattening (slope is decreasing) or growing (slope is increasing). When the concavity falls below a pre-determined level, the solution is stopped.
The fluctuation of the dependent variable about the mean value is examined. In effect, this is a measure of the standard deviation. When the fluctuation or deviation is below the set level, the solution is stopped.