Creative-GNN NACA0012 2D Compressible Flow Dataset and Reproducibility Package

This repository provides the Creative-GNN NACA0012 2D compressible-flow dataset and reproducibility package used in the associated Journal of Computational Physics submission.

Scope and physics setting:
- Geometry: NACA0012, 2D, fixed fine unstructured mesh.
- Solver/data source: SU2-based steady RANS workflow.
- Condition coverage: compressible range Ma 0.8–4.0 (archived dataset range).
- Main paper evaluation: compressible, supersonic-focused subset Ma 1.3–4.0.

Contents:
- Graph-ready processed CFD data (node/edge tables, case metadata, and boundary-related files).
- Case manifests and split definitions for controlled train/validation/test evaluation.
- Reproducibility artifacts for the reported experiments, including aggregated metrics and figure-generation inputs for field error, pressure/Cp comparisons, near-wall vs global diagnostics, shock-related diagnostics, and pressure-only force-coefficient error summaries.
- Creative-GNN code package for training/evaluation and post-processing scripts.
- Configuration files and run manifests to trace preprocessing, training, and evaluation settings.

Reproducibility goal:
This package is organized so that readers can reproduce the manuscript’s main quantitative results and figures from the archived processed data and scripts, without reconstructing the full CFD pipeline from scratch.

Repository curation note:
To satisfy repository size constraints, this version is a curated reproducibility release. It includes the files required for the published results; optional large add-ons (if any) are listed explicitly in the package documentation.

Citation:
Deng, Yuxuan (2026), “Creative-GNN NACA0012 2D Compressible Flow Dataset and Reproducibility Package”, Mendeley Data, V1, doi:10.17632/g3x4phfr3r.2