Surface Postprocessing

Some Surface postprocessing actions are available to you to either get a quick insight and indicators on the performance of your model or to download relevant outputs for further postprocessing outside of the Ansys SimAI platform.

Generate Surface VTP

The Surface VTP generation option allows you to export the geometry with the surface variables selected as "Model Output" during the model configuration. For more information, see Variables.

You can always generate a Surface VTP file, for each of the predictions you ran on the platform, with all field projected at cells (Generate at cells option).

If some fields are located at points in your training data, you can generate a Surface VTP file with fields at points and at cells, as learned by the model (Generate as learnt option).

Once the file is generated, it can be downloaded as a .vtp file.

The exported VTP file can be directly loaded in postprocessing tools such as Ansys EnSight software to compare, script and treat the prediction as a regular simulation.

When exporting the surface VTP file, the prediction data are mapped onto the elements of the surface mesh. Hence, the coarseness of this mesh will affect the resolution of the exported prediction.
Table 1. Illustrations of the impact of the coarseness of a mesh on the resolution of the exported prediction




When all fields are generated at cells, they are mesh agnostic in the sense that the field integral does not depend on mesh coarseness. It is not the case when fields are predicted at points because the values depend entirely on the positions of the mesh nodes.

Generate at cells

This option is useful:

  • If the simulation data used to build your AI model only contains fields at cells.
  • If it does not matter to your use case if the fields in the surface VTP file generated are projected at cells, even if your data are at points.

It is available through the button Generate > Generate at cells, and the generated file name will end with _surface.vtp

You can also directly preview the surface prediction on the platform by using the embedded 3D visualizer.

Generate as learnt

If you uploaded simulation data at points and/or at cells, your AI model was then trained with both. In that case, you may want to generate, export and load in postprocessing tools a file with the same characteristics, to compare what has been predicted with what you used as training data.

It is available through the button Generate > Generate as learnt, and the generated file name will end with *_surface_td_location.vtp.

The 3D visualizer is not available for that option.

Surface Evolution

The Surface Evolution postprocessing allows you to analyze how a behavior evolves along a geometry by slicing a surface into parts and computing Global Coefficients for each of these parts.

To slice your surface into several parts, you need to define only two parameters:
  • a direction → axis.
  • a step → delta.
    Note: The unit of delta is implicitly dependent on your dataset. For example, if the values in the dataset implicitly represent meters, the delta units will also be in meters.

These parameters determine how the surface is divided, allowing for a detailed evaluation of changes across different sections of your geometry. Once the surface is sliced, global coefficients derived from your surface fields are integrated using the center of each slice. The computed values are then provided in the form of downloadable files (JSON, CSV or XLSX) that can be plotted locally as evolution curves (graphs plotting the Global Coefficient against the geometry distance). The generated metrics can provide insight into local solution changes.

A practical example of this postprocessing application in the automotive industry is analyzing the coefficient of friction, which measures the resistance between tires and the road.