When using the Kriging auto-refinement capabilities, there are four main steps to the process:
Setting up the response surface
Setting up the output parameters
Updating the response surface
Generating verification points
To set up a response surface for Kriging auto-refinement:
In the Outline view for the response surface, select the Response Surface cell.
In the Meta Model section of the Properties view, set Response Surface Type to .
In the Refinement section of the Properties view:
Set Refinement Type to .
For Maximum Number of Refinement Points, enter the maximum number of refinement points that can be generated.
For Maximum Predicted Relative Error (%), enter the maximum predicted relative error that is acceptable for all parameters.
Note: For advanced options to be available, the Show Advanced Options check box must be selected on the Design Exploration tab in the Options window. For more information, see Design Exploration Options.
If advanced options are shown, in the Refinement section of the Properties view:
For Output Variable Combinations, select a value for determining how output variables are considered in terms of predicted relative error. This value controls the number of refinement points that are created per iteration.
For Crowding Distance Separation Percentage, enter a value for determining the minimum allowable distance between new refinement points.
During the refinement process, if one or more design points fail, auto-refinement uses the value specified for this property to avoid areas close to failed design points when generating the next refinement points. This property specifies the minimum allowable distance between new refinement points, providing a radius around failed design points that serves as a constraint for refinement points.
In the Verification Points section of the Properties view:
If you want to generate verification points, select the Generate
Verification Points check box. Number of Verification
Points becomes available. The default value is 1.
If you want a different number of verification points to be generated, enter the desired value.
For more information, see Kriging Auto-Refinement Properties.
In the Outline view for the response surface, select an output parameter.
In the Refinement section of the Properties view for this output parameter, select or clear the Inherit From Model Settings check box. This determines whether the maximum predicted relative error defined at the model level is applicable to the parameter.
If the Inherit From Model Settings check box is cleared, Maximum Predicted Relative Error becomes available.
If Maximum Predicted Relative Error is available, enter the maximum predicted relative error that you find acceptable. This can be different than the maximum predicted relative error defined at the model level.
For more information, see Kriging Auto-Refinement Properties.
Once you've defined the response surface and output parameter properties, complete the refinement by right-clicking the Response Surface node and selecting . Alternatively, click the Update button on the toolbar.
The generated points for the refinement appear in the refinement points table, which displays in the Table view when either the Refinement node or one of its Refinement Points subnodes is selected in the Outline view. As the refinement points are updated, the Convergence Curves chart updates dynamically, allowing you to monitor the progress of the Kriging auto-refinement. For more information, see Kriging Convergence Curves Chart.
The auto-refinement process continues until either the maximum number of refinement points is reached or the response surface is accurate enough for direct output parameters. If all output parameters have a predicted relative error that is less than the Maximum Predicted Relative Error defined for them in the Properties view, the refinement is converged.
Once the auto-refinement has converged (and particularly at the first iteration without refinement points), you should generate verification points to validate the response surface accuracy.
As with Sparse Grid, Kriging is an interpolation. Goodness of Fit is not a reliable measure for Kriging because the response surface passes through all of the design points, making the Goodness of Fit appear to be perfect. As such, the generation of verification points is essential for assessing the quality of the response surface and understanding the actual Goodness of Fit.
If the accuracy of the verification points is larger than the predicted relative error given by Kriging, you can insert the verification points as refinement points and then run a new auto-refinement so that the new points are included in the generation of the response surface. Verification points can only be inserted as refinement points in manual refinement mode. For more information, see Verification Points.