Geometries#

Geometries are the 3D meshes representing the surface of your design. Running predictions on geometries will output simulation results based on the trained AI model.

Supported File Formats#

The supported file formats are VTP (recommended) or STL:

  • VTP should be used whenever possible. It will be faster to process by the application.

Important

This format is mandatory if you intend to use Surface fields as input variables for prediction, since STL format does not handle surface variables.

  • STL can be used as an alternative, although heavier and slower.

Important

(For all formats) If normals are provided, but contain NaN, those will be interpreted as 0 for postprocessing and left as NaN for download (off-platform postprocessing, etc.). Centroids and area will always be computed by the platform. Other fields will be kept as submitted if not overwritten by the prediction values.

Sanity Checks before Upload#

Perform the following sanity checks to your prediction geometries to ensure an optimal usage of the application’s prediction capabilities:

  • Geometry length: Check for geometry length before adding. A model can tolerate geometry lengths (on x-axis) between 80% of the smallest and 120% of the longest geometries seen during its training.

  • Normal orientation: Normals should be oriented consistently, either all towards the exterior, or towards the interior. This consistency should be kept for both AI-model creation and prediction.

  • Overlapping surface: Use geometric algorithms (bounding box checks or mesh intersection tests) to identify and correct potential overlapping regions between surfaces to ensure the data do not contain duplicate or conflicting information.