{"success":true,"data":{"author":"Jessica Ezemba, Carnegie Mellon University Design Research Collective","bucketAccessType":"public","bucketName":"cmudrc/SFEM","bucketPrefix":"","category":"Physics","cloudProvider":"huggingface","createdAt":"2026-11-18T00:06:05.772Z","description":"Three-dimensional finite element meshes with stress and displacement fields under stochastic point loading, built at the Carnegie Mellon University Design Research Collective for neural network surrogates of the stochastic finite element method (SFEM). About 16,000 CAD geometries generated with BrepGen were meshed with Gmsh (Frontal-Delaunay tetrahedra) and solved with FEniCSx/DOLFINx as linear elastic parts (E = 2.303 GPa, nu = 0.4002) under point loads of 200 N, 2,000 N and 20,000 N, with 50 realizations per geometry and load. Per sample fields include vertices, von Mises stress, displacement, the full stress tensor, load class and a fixed-facet boundary mask. Step_Files/ holds the original STEP geometries (about 39,900 files, 3 GB); Pytorch_Geometric_Files/ holds processed PyTorch Geometric train and validation sets (about 7 GB and 1.2 GB) and raw tar.gz archives, split 85/15 by geometry. Total about 18.6 GB. Curated by Jessica Ezemba; licensed MIT as stated in the card; cite Neural Network Surrogate Modeling for Stochastic FEM using 3D Graph Representations (2025).","displayTitle":"SFEM Dataset: 3D Finite Element Stress Fields under Stochastic Point Loads for Graph Neural Network Surrogates","huggingFaceUrl":"https://huggingface.co/datasets/cmudrc/SFEM","id":"e27f1fa1-4124-54d3-be9b-6fbcedb93be5","license":"MIT","licenseEvidenceUrl":"https://huggingface.co/datasets/cmudrc/SFEM","objectCountEstimate":39902,"probeStatus":"ok","referenceUrl":"https://github.com/cmudrc/SFEM","region":"us","schemaCard":{"accessHints":[{"language":"python-datasets","snippet":"from datasets import load_dataset\nds = load_dataset(\"cmudrc/SFEM\")"},{"language":"python-hub","snippet":"from huggingface_hub import snapshot_download\nsnapshot_download(repo_id=\"cmudrc/SFEM\", repo_type=\"dataset\")"}],"caveats":["HuggingFace metadata: author=cmudrc, license:mit, size_categories:10B<n<100B, downloads=1301"],"format":"mixed","generatedAt":"2026-11-18T00:09:54.472Z","layout":{"example":"Pytorch_Geometric_Files/processed/pre_filter.pt","pattern":"{repoId}/{filename}"},"mountPath":"/mnt/archive","sampleHeaders":[{"bytes":"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","encoding":"hex","key":"Pytorch_Geometric_Files/processed/pre_filter.pt","truncatedAt":512}],"sampleObjects":["Pytorch_Geometric_Files/processed/pre_filter.pt","Pytorch_Geometric_Files/processed/pre_transform.pt","Pytorch_Geometric_Files/processed/train_data.pt","Pytorch_Geometric_Files/processed/val_data.pt","Pytorch_Geometric_Files/raw_train_data.tar.gz"],"sidecars":[{"bytes":"---\nlicense: mit\ntask_categories:\n- other\nlanguage:\n- en\ntags:\n- finite-element-method\n- stochastic-simulation\n- graph-neural-networks\n- engineering\n- mechanical-engineering\n- stress-analysis\n- pytorch-geometric\nsize_categories:\n- 10B<n<100B\npretty_name: SFEM Dataset - Neural Network Surrogate Modeling\n---\n\n# Dataset Card for SFEM Dataset\n\nThis dataset provides 3D finite element meshes with corresponding stress distributions under stochastic point elastic loading conditions, designed for training neural network surrogate models for the stochastic finite element method (SFEM).\n\n## Dataset Details\n\n### Dataset Description\n\nThis dataset enables researchers to develop graph neural networks that can predict stress distributions orders of magnitude faster than traditional SFEM simulations. It contains approximately 16,000 CAD geometries with corresponding finite element analysis results under various loading conditions.\n\n- **Curated by:** Jessica Ezemba\n- **Language(s):** English (documentation)\n- **License:** MIT\n\n### Dataset Sources\n\n- **Repository:** https://github.com/cmudrc/SFEM\n- **Paper:** Neural Network Surrogate Modeling for Stochastic FEM using 3D Graph Representations: A Comparative Study (2025)\n\n## Uses\n\n### Direct Use\n\nThis dataset is intended for:\n- Training graph neural networks for engineering design applications\n- Developing surrogate models for finite element analysis\n- Research in computational mechanics and machine learning\n- Educational purposes in engineering simulation\n\n### Out-of-Scope Use\n\nThis dataset should not be used for:\n- Safety-critical applications without proper validation\n- Real-world structural analysis without engineering oversight\n- Applications requiring non-linear material behavior\n\n## Dataset Structure\n\nThe dataset follows PyTorch Geometric conventions:\n\n```\nDataset/\n├── Pytorch_Geometric_Files/\n│   ├── processed/\n│   │   ├── pre_filter.pt       # Preprocessing filters\n│   │   ├── pre_transform.pt    # Data transformations  \n│   │   ├── train_data.pt       # Processed training data (~7GB)\n│   │   └── val_data.pt         # Processed validation data (~1.2GB)\n│   ├── raw_train_data.tar.gz   # Archived raw training files\n│   └── raw_val_data.tar.gz     # Archived raw validation files\n└── Step_Files/                 # Original CAD geometries (3GB)\n    ├── 000/                    # STEP files (subset 1)\n    ├── 001/                    # STEP files (subset 2)\n    └── 002/                    # STEP files (subset 3)\n```\n\n**Data Fields:**\n- `vertices`: 3D coordinates of mesh vertices (N × 3)\n- `von_mises_stress`: Stress values at each vertex (N × 1)\n- `displacement`: Nodal displacement vectors (N × 3)\n- `stress_tensor`: Full stress tensor components (N × 3 × 3)\n- `load_class`: Loading condition (small_Load, medium_Load, large_Load)\n- `fixed_facet_mask`: Boundary condition indicators\n\n**Data Splits:**\n- Training: 85% of geometries\n- Validation: 15% of geometries (geometry-based split for proper generalization)\n\n## Dataset Creation\n\n### Curation Rationale\n\nTraditional stochastic finite element methods require thousands of deterministic FEM evaluations, creating prohibitive computational costs. This dataset was created to enable neural network surrogates that can predict stress distributions in milliseconds rather than hours, supporting efficient iterative design exploration.\n\n### Source Data\n\n#### Data Collection and Processing\n\n- **Geometry Source:** CAD geometries generated using [BrepGen](https://github.com/samxuxiang/BrepGen)\n- **Mesh Generation:** Gmsh with Frontal-Delaunay algorithm for tetrahedral meshes\n- **FEM Software:** FEniCSx (DOLFINx) for finite element analysis\n- **Material Properties:** Linear elastic (E = 2.303 GPa, ν = 0.4002)\n- **Loading Conditions:** Stochastic point loads (200N, 2000N, 20000N)\n- **Statistical Sampling:** 50 realizations per geometry/load combination\n\n#### Who are the source data producers?\n\nThe simulation data was generated using open-source computational tools:\n","encoding":"utf8","key":"README.md","name":"README.md","size":5460,"truncated":true}],"sourceObjectKey":"Pytorch_Geometric_Files/processed/pre_filter.pt","starterCell":"from huggingface_hub import hf_hub_download\npath = hf_hub_download(repo_id=\"cmudrc/SFEM\", filename=\"Pytorch_Geometric_Files/processed/pre_filter.pt\", repo_type=\"dataset\")\nprint(path)","starterSamplePath":"Pytorch_Geometric_Files/processed/pre_filter.pt","totalObjects":39902},"summary":"Three-dimensional finite element meshes with stress and displacement fields under stochastic point loading, built at the Carnegie Mellon University Design Research Collective for…","title":"SFEM Dataset: 3D Finite Element Stress Fields under Stochastic Point Loads for Graph Neural Network Surrogates","updatedAt":"2026-11-18T03:54:20.558Z","visibility":"public","websiteUrl":"https://huggingface.co/datasets/cmudrc/SFEM"}}