PLAID Datasets¶

At the heart of PLAID is the seemless exchange of dataset for streamlined collaboration. Corresponding datasets are provided at HuggingFace and in a Zenodo community.

PLAID‑datasets is a curated collection of physics‑based datasets formatted with the PLAID data model. Each dataset represents a computational physics or structural mechanics problem defined on unstructured meshes and is suitable for graph machine‑learning applications. The datasets contain mesh data together with scalar and field quantities (e.g., pressure, displacement and turbulence variables) and come with predefined train/test splits to support machine‑learning tasks. They were all generated using the PLAID library and datamodel and are released under open licences. See also PLAID's official documentation for more information.

Original datasets¶

Tensile2d

Visualization appli:

2D_MultiScHypEl

Visualization appli:

2D_ElPlDynamics

Visualization appli:

Rotor37

Visualization appli:

2D_profile

Visualization appli:

VKI-LS59

Visualization appli:

AirfRANS datasets¶

AirfRANS, introduced in ¹ is an additional dataset provided in PLAID format and various variants. Since the outputs on the testing sets are public, no benchmark application is provided for this dataset and its variants.

Dataset image

AirfRANS (original)

Visualization appli:

AirfRANS (clipped)

Visualization appli:

AirfRANS (remeshed)

Visualization appli:

References¶

F. Bonnet, J. Mazari, P. Cinnella, and P. Gallinari. AirfRANS: High Fidelity Computational Fluid Dynamics Dataset for Approximating Reynolds-Averaged Navier–Stokes Solutions. In S. Koyejo, S. Mohamed, A. Agarwal, D. Belgrave, K. Cho, and A. Oh, editors, Advances in Neural Information Processing Systems, volume 35, 23463–23478. Curran Associates, Inc., 2022. URL: https://arxiv.org/abs/2212.07564. ↩