The Uncertainty Quantification (
) node under
Job Configurations is used by the UQ study. Most of the settings are shared and synchronized from the study level to the Job Configuration. Another related node is the Design of Experiment Job Configuration, which also shares many of its settings with the study. The organization of the settings is very much the same. The differences between the nodes are mentioned below.
The synchronization from study is dictated by the Defined by study step setting at the top of the
General section. When this setting points to a UQ study, all synchronized properties are grayed out and cannot be edited. By changing this setting to
User defined, you can change these and run the UQ analysis directly from the Job Configuration.
This section also has a setting for which Design of Experiments job to use. By default, this points to
New and a node of type
Design of Experiments is added. You can also use an existing node of this type.
For the SPCE and ASPCE surrogate models, you can find a Maximum polynomial degree setting (default value: 30), which terminates the increase of order for the SPCE construction.
For the GP and AGP surrogate models, you can find three settings for the training of the GP. These settings are related to the so-called hyperparameters, which can be seen as an internal tuning of the GP to a fixed dataset. These are the Method for the number of restart points for training list:
Automatic (the default) and
Manual. The
Automatic method will use
40m restart points. For each of these points, a local optimization problem is solved to find the best parameters. With the
Manual method, specify the
Number of restart points for training. With the
Relative tolerance for training, you can set the termination tolerance for the local optimizer.
There are two Latin hypercube sampling settings that are used by the used Design of Experiments node: the
Maximum number of iterations for LHS and
Number of restart points for LHS. These settings are related to the problem of finding an optimal set of sampling points. See
Design of Experiments for more information.
The checkbox Distribute parameters is synchronized to the
Distribute model evaluation setting in the
Advanced section setting for
The Uncertainty Quantification Study. There is also a
Maximum number of groups setting, which is not used when synchronized to a UQ study. When in
User defined mode, you can change this and prescribe a value. Groups can be used to split all the evaluations into groups, so that each cluster node evaluates a group of model evaluations together.
The main setting in this section is the Error handling list. It is synchronized to the setting in the
Advanced section with the same name for
The Uncertainty Quantification Study. All warnings and errors thrown within an uncertainty quantification study are collected in the table in this section. If errors occur for the
Skip problematic parameters method, the failing parameter values and the corresponding error messages are both collected in the table. When there are warning or error messages collected, this node is also decorated with a Warning overlay on its icon.
) and select Solver and Job Configurations in the Show More Options dialog.