Skip to main content

AI Physics

The Rescale AI Physics Operating System (OS) provides a guided, end-to-end way to turn simulation data into high-accuracy surrogate models with fast and flexible inference. It is designed for simulation engineers (not AI experts) and supports a range of production-ready model architectures across multiple solvers.

A surrogate model is a fast approximation of a simulation solver, trained on completed simulation results. Once trained, a surrogate model can predict outputs for new inputs in seconds instead of hours, allowing you to explore more design candidates without running additional solver jobs.

Benefits​

Accelerate design exploration Run thousands of design iterations in the time it would take to complete a handful of full simulations. Surrogate models predict results in seconds, enabling rapid what-if analysis and sensitivity studies.

Reduce compute costs Train once on a representative sample of your design space, then run unlimited predictions at a fraction of the cost of full solver runs. Ideal for optimization loops and Monte Carlo studies.

Enable real-time decision making Integrate surrogate predictions into CAD tools, optimization frameworks, and custom workflows. Get instant feedback on design changes without waiting for solver jobs to complete.

No AI expertise required Automated hyperparameter tuning (Auto Tune), guided case validation, and pre-configured model architectures mean you don't need machine learning experience to build production-quality surrogate models.

The AI Physics Workflow​

AI Physics follows a four-stage workflow:

1. Prepare Simulation Jobs​

Run a parameter sweep or design of experiments (DOE) covering the input space you want the surrogate model to predict over. Enable Metadata Extraction and AI Extraction Automations on your Rescale jobs so results are automatically prepared for AI training.

Output: A set of completed simulation jobs with extracted inputs and outputs.

→ Learn more: Launching a training session

2. Launch a Training Session​

Open a Training Session to get AI Physics Training, the interface where you create a model, point it at your completed jobs, and validate that the cases are consistent. Data selection happens inside model creation rather than as a separate dataset-building step.

Output: A Training Session running AI Physics Training, with validated cases ready to configure.

→ Learn more: Launching a training session

→ Reference: Working with AI datasets, for the notebook path and for merging, splitting and inspecting a dataset afterwards

3. Train a Surrogate Model​

Select a model architecture (DoMINO, TwoStreamTransformer, GeoTransolver, or MeshGraphNet) and configure training. Auto Tune (recommended) automatically explores hyperparameter configurations in parallel and selects the best model. Manual mode gives you full control over every training parameter.

Output: A trained surrogate model with evaluation metrics (R² score, residual distributions, error visualizations).

→ Learn more: Training Surrogate Models

4. Run Inference​

Use your trained model to predict simulation outputs for new input configurations: in AI Physics Training, in AI Physics Inference, locally on your workstation, or directly in CAD tools (Blender, Autodesk Alias).

Output: Instant predictions enabling rapid design exploration.

→ Learn more: Running Inference

Prerequisites​

Before you begin:

  • AI Physics must be enabled for your organization. If you don't see Model Builder or AI Models in the Rescale side navigation, contact your Rescale representative.
  • Create an API key from your User Profile Settings to enable AI Model creation in the Rescale platform.
  • Have completed simulation jobs from a parameter sweep or DOE with Metadata Extraction and AI Extraction Automations enabled.

Supported solvers: AI Physics supports a growing list of CFD and FEA solvers. Contact your Rescale representative for the full list of supported solvers and to request support for additional solvers.

Tutorials​

Follow the tutorials below to train your first surrogate model end-to-end. Each tutorial uses pre-prepared job results so you can walk through the full AI Physics workflow without setting up your own DOE first. Each one covers case selection, model training and inference.

Get Started​

Ready to build your first surrogate model?

  1. Launch a training session: prepare your simulation jobs and open AI Physics Training
  2. Train a Surrogate Model: use Auto Tune to build a high-accuracy surrogate model
  3. Run Inference: predict outputs for new designs in seconds
  4. Explore the design space: sweep and optimize across hundreds of design variants

FAQs​

What is a surrogate model, and why would I use one? A surrogate model is trained on completed simulation results. Once trained, it can predict outputs for new inputs in seconds rather than hours. Surrogate models are useful for design space exploration, sensitivity analysis, and optimization studies where running a full solver for every configuration would be too expensive or slow.

Do I need AI or machine learning experience to use AI Physics? No. AI Physics is designed for simulation engineers. The Jupyter Notebook templates, default model parameters, and built-in evaluation dashboards guide you through each step. An AI copilot is also available in the notebook interface to help answer questions and suggest next steps using natural language.

How do I know if my trained model is accurate enough to use? After training, review the evaluation dashboards in AI Physics Training. These show R² scores broken down by output variable, residual distributions, and 3D error visualizations across your test cases. A high R² (close to 1.0) and low residuals across your test set indicate the model generalizes well. Always validate a sample of predictions against full solver runs before using the model for critical decisions.

Can I retrain or fine-tune a model after the initial training run? Yes. In AI Physics Training, you can create a new model version that inherits the prior model's architecture and configuration as a starting point. With Auto Tune, use the Optional Hyperparameter Seed to seed from the prior model's hyperparameters. This is useful when you want to improve accuracy without starting from scratch, or when you have new training data to add.

What if I'm using a solver that isn't listed as supported? AI Physics requires Metadata Extraction and AI Extraction Automations to be available for your solver. If your solver is not yet supported, contact your Rescale representative. Solver support is expanding over time, and your use case may inform the roadmap.

How much does AI Physics cost? AI Physics training runs on GPU hardware charged by core-hour. Auto Tune search costs are bounded by: search duration × hardware hourly rate. For example, a 6-hour Baseline Search on Grossular-4 ($6.87/hr) costs at most ~$41. Inference costs depend on your chosen environment. AI Physics Inference runs on Rescale infrastructure (charged by inference job), while local inference runs on your own hardware at no additional cost.

Support​

For questions, support, or to request additional solver support, contact your Rescale representative or visit the Rescale Support Center.