Use the following best practices to evaluate the convergence and the accuracy of data transfers:
- Set an appropriate convergence target.
Set data transfer and participant convergence target that are strict enough to ensure that the data on the target is not affected by the input data once the convergence criteria have been reached.
- Review convergence diagnostics.
Review convergence diagnostics for each coupling iteration of each coupling step. The coupling step/iteration sections of the Transcript/Log show the convergence status for data transfers and participants. Data transfers, and, in most cases, participants should reach convergence in each coupling step.
Data transfer diagnostics:
Review the data transfer diagnostics reported at analysis initialization and for each coupling iteration of each coupling step. The coupling step/iteration sections of the Transcript/Log show the values data transfer values reported by the source and target participants. The Sum is reported for extensive variables (mapped using a conservative algorithm) and the Weighted Average (over all regions in the interface) is reported for intensive variables (mapped using a profile-preserving algorithm).
The reported values should be the same for both the source and the target of the data transfer. When they are not:
If the target sum is less than the source sum, then this indicates (and should confirm) that System Coupling is implementing a method of generating data on unmapped target locations according to the type of mapping used for the transfer.
For more information, see Generating Data for Target Locations.
If the target weighted average significantly differs from the source weighted average, then the mapped results are suspect and should be reviewed in greater detail. This can often be attributed to the target-side meshes that are coarse relative to the source-side meshes, especially in regions with significant source-solution variations.
Participant diagnostics:
The participant-solution status section indicates whether each participant completed its step during the iteration.