The Field data models menu option creates a random field model based on simple numerical autocorrelation models that have the property that each scatter shape contains a single maximum. This maximum is located at a support point (similar to RBF networks). Neighbouring support points are automatically collected into a single shape. Thus, you can manually define scatter shapes not only around points but also around edges or surface patches.
Note: For the cautions that apply to beta features, see the introduction in the optiSLang Beta Features guide.
To create a free-form variation model, you set these options in the dialog box that opens:
- Ident
The unique identifier for the new quantity.
- Data type
Data type of the new free-form variation model. Choices are Element or Node.
- Number of auto-parameters
Support points to find scatter shapes, placed by the algorithm. A higher number of auto-parameters shortens the autocorrelation length and generally leads to the generation of a higher number of more localized scatter shapes. Manual support points must be specified if set to zero.
- Use manual support points
If this check box is selected, the algorithm is to incorporate the support points from a manually created component, in addition to the number of auto-parameters. The manual support points must match the data type selection (Element or Node).
- Magnitude of variation
Create a free-form variation model with homogeneous variation at every mesh point or use a the standard deviation given by a data object, matching the data type selection (Element or Node).
- Show advanced options
If this check box is selected, these advanced options display.
Normalized shape of variation patterns
the shapes are scaled such that their sum equals the standard deviation. If this check box is cleared, the standard deviation is the upper bound to the sum of the shapes.
Use mesh distance
If this check box is selected, distances are computed along the mesh edges. If this check box is cleared, the Cartesian coordinates are used to compute the distance between two points.
Use mean as parameter
If this check box is selected, the mean is associated with a scaling parameter and is the first shape. Use this option to scale a field property proportional to the mean. You cannot require this option if normalized shapes are used because you then might have the same effect when setting all random field scaling parameters constant.
Compute individual variations
If this check box is selected, the individual variation (explainable variance by a single scatter shape) are computed as additional data objects.
Erase degenerated shapes
If this check box is selected, automatically created variation patterns having a short correlation length are removed from the model.
Compute distance object
If this check box is selected, the distances between the support points are computed and additional data objects are created. One object is created for each shape, representing the distance field around its support point and a distance field indicating the closest distance between all support points. This can help to debug issues with automatically generated support points.
Enforce zero variation at boundary
If this check box is selected, oSP3D tries to enforce a near-zero value at the boundary of the currently active set. Use this option if the active set lies directly next to a fixed region.
Note: Enforcing zero variation at the boundary does not always work perfectly and requires normalized shapes.
Boundary to be used for automatic support
Distance calculations for automatic support point placement require oSP3D to identify boundaries. Choices are:
Default: oSP3D selects an appropriate mode.
None: When using manual support points only, no boundary is required.
Nodes on surface and on reference set boundary: oSP3D identifies free-floating boundaries.
Nodes on boundary edges: Use with shell-like meshes have a closed boundary.
Mean Value
Create a free-form variation model with a homogeneous mean value at every mesh point or use a the mean given by a data object, matching the data type selection (Element or Node).
Output Objects for Each Quantity
For each quantity, these output objects are generated:
Random field model
Individual scatter shapes as results in the data object table:
shape[1], …, andshape[. The data object table shows the cumulative explainable variation for the complete field (as average) for each shape.n]Explainable variation of the random field model as a result in the data object table:
variation. The data object table shows the explainable variation for the complete field (as average).Distance objects:
For each shape, one distance object named
shape_distance[n].One object named
shape_distance_supports, indicating the distance between closest support points.
Log messages: oSP3D determines the fraction of explainable variation that is obtained for individual numbers of scatter shapes and prints this table in the log messages.
Usage
The free-form variation model offers the possibility of building a random field model even if insufficient data is available to estimate the true correlation structure, with the additional option of centering scatter shapes on points of particular interest using manual support points.
The mean and standard deviation from a small number or incomplete measurements can improve the free-form variation model.