Crash-GNN deformation prediction datasets

This dataset contains the finite-element simulation outputs used in the article “A Non-Intrusive Graph Neural Network Surrogate for Crash Deformation Prediction”.

The data consist of transient deformation histories generated from parametric explicit finite-element simulations of a rigid impactor striking a clamped thin aluminium plate. The files are provided as compressed NumPy files (.npz) and are organised into the same dataset splits used in the paper.

The repository includes the following datasets:

- BallSize: variation of spherical impactor radius.- BallPosition: variation of impact location with fixed spherical impactor radius.- BallImpact: joint variation of impactor radius, impact location, and incidence angle.- geometric_transfer/cubic_impactor: cubic impactor cases used for geometric-transfer evaluation.

Each file includes the information required to construct the graph-based representation used in the paper, including static mesh information, dynamic nodal fields and boundary-condition.