Home
Design Exploration
Once the baseline model is prepared, you can define morph volumes, morph geometry, create design shapes and run DOE studies. These tool work for both AcuSolve and ultraFluidX-based workflows.
PhysicsAI
PhysicsAI Capabilities in CFD
PhysicsAI is an AI-powered tool designed to accelerate simulation workflows by leveraging historical simulation data.

Welcome
What's New
View new features for HyperMesh CFD 2025.1.
Get Started
Learn the basics and discover the workspace.
Onboarding Resources
Discover HyperMesh CFD functionality with interactive tutorials and community resources.
Solution Layers
Solution-centric workflows allow you to traverse through modeling environments using a dedicated UI layer.
Manage Files and Data
Create, open, import, and save models.
Convert
Use the Convert tool to convert entire models between data types.
Geometry Repair
Manage CAD, FE, and 2D parametric sketch geometry.
Meshing
Generate surface/volume mesh by defining mesh controls, or interactively create and edit 2D surface mesh.
Setup
Prepare your model and run the simulation.
Post-Processing
Post-process the simulation results by creating visualizations and measurements.
Design Exploration
Once the baseline model is prepared, you can define morph volumes, morph geometry, create design shapes and run DOE studies. These tool work for both AcuSolve and ultraFluidX-based workflows.
- Morphing
- Morph
  Rapidly change the shape of the FE mesh without severely sacrificing the mesh quality.
- PhysicsAI
  - PhysicsAI Capabilities in CFD
    PhysicsAI is an AI-powered tool designed to accelerate simulation workflows by leveraging historical simulation data.
    - Decimate
      The Decimate tool reduces the number of mesh elements while preserving key features of the geometry, simplifying complex models without sacrificing accuracy.
  - Guidelines for PhysicsAI Usage
  - Create and Load Projects
    Each session starts with creating a new project or loading an existing project.
  - Create and Manage Datasets
    Datasets are collections of one or more simulation results used for training or testing a model.
  - Train Models
    Train models using any physics, any mesh, and without design variable parametrization.
  - Test Models
    Use the Model Testing tools to test models and review test results.
  - Predict Results
    Use the predict tool to generate a model prediction.
  - Similarity Score
    Like all machine learning models, PhysicsAI models are most accurate when the design being predicted is similar to the designs used for training.
  - Use Batch Mode
    Learn more about using batch mode.
  - Best Practices for PhysicsAI
    Learn more about the best practices for training and using PhysicsAI models.
  - Frequently Asked Questions
    This section provides quick responses to typical and frequently asked questions regarding PhysicsAI.
Automation
Automate simulation processes with templates and utility scripts.
Solver Help
More resources for AcuSolve and ultraFluidX solvers.

View All Altair HyperWorks Help

PhysicsAI Capabilities in CFD

PhysicsAI is an AI-powered tool designed to accelerate simulation workflows by leveraging historical simulation data.

Unlike traditional solvers that require extensive meshing and parameterization, PhysicsAI employs geometric deep learning to learn relationships between geometry and physics outcomes directly from CAD or mesh models. This approach enables rapid predictions, facilitating quicker design iterations and decision-making.

KPIs (Key Performance Indicators)

KPIs are quantifiable metrics that help to assess the accuracy, efficiency, and effectiveness of simulations.

Useful to predict quantities; shows very promising results.
No need to decimate result data.
Fast to train.
No heavy compute resources need to train.

Field Prediction

Field prediction is the estimation of field variables, like velocity, pressure, temperature, or turbulence quantities, across the entire computational domain of a fluid simulation, without solving the full set of CFD equations.
Useful for shape exploration with consistent set of training data:
- Same vehicle configuration (sedan, hatch back, SUV, etc.)
- Same physics conditions (condition, fan condition, tire rotation)
- No porous media; no baffles
Decimation is required for optimal results on A100 GPU. We are working to provide guidelines about ideal node count for available compute resources.
CFD is able to decimate PowerFlow, Fluent, and Star-CCM+ data (given the ensight file format)
Decimation does have impact on accuracy (1-5 % for CD). For example, in the DriVaer model reducing the nodes by 60% shows a 0.5% error in Cd*A. Resulting contours look comparable.