COMSOL 6.4 - Parametric Sweep (Job Configurations)

) job configuration, which can loop over a given set of parameters. For each set of parameters, it runs the sequence defined by its subnodes. You can combine the sequence with other Batch, Parametric, or Cluster Computing sequences in a hierarchical way by adding a job configuration that points to another node. You can, for instance, create a node that does a LiveLink™ update and then runs a node that in turn runs a second sequence on another node.

A parametric sweep can contain steps to run all solutions, save the model to a file, and evaluate derived values, for example. You add those steps by right-clicking the node and selecting one of the following options:

refers to another job that is to be run from this parametric sweep, while (see Solution (Subnode)) runs a node as available under the node, available further up in the tree.

Under , you can choose (see External Class (Subnode)), which calls an external Java® class file. Another option, (see Method Call (Subnode)), runs a model method. The (see Geometry (Subnode)), builds the node (runs the geometry sequence). This can be used, for example, in combination with a parametric sweep to generate a sequence of MPH files with different geometry parameters. The option (see Mesh (Subnode)) builds the node (runs the meshing sequence).

Under , you can choose (see Plot Group (Subnode)) to run all or a selected set of plot groups. This is useful for automating the generation of plot groups after solving. The option is there for legacy reason, and it is instead recommended that you use the option (see Evaluate Derived Values (Subnode)), which evaluates nodes under > . The option (see Export to File (Subnode)) runs any node for data export under the node.

Use the list to specify that the settings are synchronized with the corresponding study step.

Use the list to specify the type of sweep to perform. The type (the default) solves for a number of given combinations of values, while the type solves for all combinations of values. Using all combinations can lead to a very large number of solutions. The type makes it possible to run sweeps over selected parameter groups and associated parameter cases. For details about each sweep type and instructions on defining the parameters, see Parametric Sweep.

You can add a that is evaluated after each solution. The last solution, which breaks the stop condition, is discarded from the set of solutions returned (the node) but is included in probe table outputs. Once the condition evaluates to a negative value, the node is stopped. Select the checkbox to store both solutions — before and after the stop condition has been satisfied — in the parametric solutions storage.

See Results While Solving in the Common Study Step Settings section.

Errors are usually stored in the table. If you want to get the error message at once, select the checkbox. Select the checkbox to allow the sequence to continue even if an error is encountered. This is different from where the sweep only continues with the next step if no error occurs. The option is, for instance, useful when you are interested in the solution even if there is an error.

You can distribute the sweep on several computational nodes by selecting the checkbox. If the problem is too large to run on a single node, enable the field to use the nodes’ memory more efficiently. In this case, the same parameter is solved for by several nodes that cooperate as if running a nondistributed sweep. The number of nodes that cooperate is equal to the maximum of the total number of nodes divided by the setting and 1. So if the total number of nodes is 12 and the is 3, 3 groups with 4 nodes each cooperate.

To improve robustness of distributed parametric sweeps, the is selected by default from the list. This will insert checkpoints in the sweep at which the solutions, exported files and probe tables will be synchronized. A recovery file will also be saved at each checkpoint. Each checkpoint splits the sweep in smaller sweeps and the progress can be recovered up until the last checkpoint. Enter a positive integer in the field (default: 4) to insert a checkpoint every N parameters for each group of nodes. If the total number of groups is 3, and N is 4, a checkpoint will be inserted every 12 parameters. If the parameters influence the solving time strongly a smaller N could result in extra synchronization time if one group reaches the checkpoint faster than other groups, choosing N is a trade-off between robustness and speed. Select if no checkpoints should be inserted and the sweep should be optimized for speed.

Select the checkbox as needed.