Overview

Physics AI Foundation Models can transform automotive design by providing real-time physics-based performance feedback. However, development has been hindered by insufficient training data. Luminary SHIFT Models provide access to both high-quality datasets and pre-trained foundation models for a variety of applications.

SHIFT-SUV is built on thousands of simulations of parametrically modified AeroSUV platform variants. These simulations were performed on Luminary cloud using our industry-validated mesh generation and scale-resolving DES solver.

Visit our blog for more details on dataset creation.

What's included?

For each simulation we provide the following:

• geometry file (STL)
• surface and volume flow field (VTK)
• charts showing the time history of solver residuals and forces (PNG)
• a time history of the drag and lift forces (CSV)
• directory with flow field images of the geometry (PNG)

The pretrained foundation model is based on the NVIDIA PhysicsNeMo DoMINO architecture.

Access Shift-SUV

A sample dataset (99 simulations, no volumetric data) is available here and distributed under the open source CC-BY license (no restriction on use). 

To access the complete dataset and foundation models, please complete the form on this page and we will contact you.

Important Licensing Note

The AeroSUV platform remains subject to a non-commercial use restriction. SHIFT-SUV does not alter this restriction—usage of the AeroSUV geometry is limited to non-commercial applications.

FAQs

What license is this distributed under?
The dataset and foundation model are distributed under the CC BY-NC license, requiring attribution and restricted to non-commercial use. 

Can I use this for commercial purposes?
The dataset and model are distributed under a non-commercial license. For commercial use exceptions, please contact us using the form above to explore options.

How can I contribute to this open source project?
Please contact us to learn how to contribute to this project, or to leverage this dataset in your own project(s).

How was this model and dataset created?
The dataset was generated by parametrically morphing the AeroSUV CAD model to generate thousands of geometries and associated meshes (ANSA). These were simulated with transient, scale-resolving Detached Eddy Simulation (DES) in the Luminary platform. See our blog post for more details.