To evaluate the convergence of data transfers, each iteration is measured against the
previous iteration. The change in all the data transfer values between these two
successive iterations is reduced to a normalized value. When two successive iterations
produce a normalized value that is under the convergence (you can change the convergence
target default value of
1e-2
), the data transferred has converged.
Two global (that is, over all locations) measures of convergence are evaluated and reported during execution of the coupled analysis. These include the maximum and Root Mean Square (RMS) of the normalized change in data transfer values. The RMS is the default measure used to determine convergence. The measure can be changed to the maximum of the normalized value through the System Coupling Input file.
The RMS value is evaluated as:
 | (1) |
where 
is the normalized change in the data transfer value between successive
iterations within/across a given coupling step, and is measured as:
 | (2) |
where 
is the data transfer value.
In Equation 2, the denominator, or normalization factor, is evaluated differently in the transient and general coupling analyses. In the transient coupling case, the normalization factor equals the average of the range and mean of the magnitude of data transfer values over all locations for the current iteration. In the general coupling case, it equals the average of the range and mean of the magnitude of data transfer values over all locations for all iterations in the entire analysis. This normalization factor is a representative scale for the data transfer values and ensures that division by zero (due either to zero range or zero mean) is avoided.
In Equation 2, 
is the un-normalized change between successive iterations, and is expressed
as:
 | (3) |
where 
and 
correspond to the current and the previous iterations respectively,
and 
is the under-relaxation factor applied in forming the final value
applied during the current iteration. In the first coupling iteration of every coupling
step, 
is assumed to be unity. If the variable is a vector and/or complex,
then and 
represent the variable magnitude.
When there is no change in data transfer values, the default for
RMS/MAX is
1.0e-014.
Note: The global data transfer convergence measures are set to unity in the first coupling iteration of the first coupling step during an initial run. After a restart, if a data transfer involving a new variable is defined or if the region is remeshed, these measures are set to unity in the first coupling iteration of the first coupling step.
Although monotonic convergence to the specified target values is ideal, oscillatory convergence and/or divergence (that is, constant or increasing convergence measures) may also occur.
For best practices, see Evaluating Convergence and Data Transfer Accuracy.