Learn how to configure all relevant parameters of your AI model and how to build
it.
Prerequisites: You have created a project and added a minimum of four simulations in the Ansys SimAI platform as described in Creating a Project and Adding Data.Tip: Model
configuration is auto saved in your browser. Any edit alters the definition of your
upcoming build request. Click Revert if necessary to revert the
model configuration to its last available version, that is, the last configuration used
to build a model in this project.
-
From your project view, click Build new model.
-
In the Reference sample section, verify that the
dataset selected (to extract your variables) is valid.
If you want a different training data point to be used as reference sample by
your model, click its info button and click Set as project
reference sample. For more information, see Reference Sample.
-
In the Variables section:
-
Click Select variables to select the variables
relevant for your study.
-
Click Create coefficient to create a scalar
value (for example: Drag, Lift) that will be computed as a
postprocessing after a prediction.
-
Name the global coefficient in the Name field
and click Create Coefficient.
-
Click Define domain of analysis to determine the size
and position of the domain (bounding box) used for AI learning and model
predictions:
- Use the Quick setup to automatically get the
domain extends based on all the geometries in the project dataset by
clicking Apply quick setup and
Validate.
- Use Custom setup to manually define the domain
extends and click Validate.
Tip:
The relative positions (either to the design edge or the design center)
allow the
Ansys SimAI platform to adapt the training and
prediction bounding box to the geometries' position variation (for
translation).
Make sure the relative settings generate a DoA that is consistently
positioned for all geometrical variations of the designs.
-
In the Build Duration section, choose the right
accuracy for your analysis:
-
Click Build Model.
The AI training is started. You will receive an email when your AI model is
ready.
Once your AI model built in production, if you need to
build another model with the same configurations, use the Build on
top feature. It reduces the build duration. For more information, see Building on Top of Previous Model (Deprecated).