---
license: cc-by-nc-4.0
language:
- zh
- en
pretty_name: HVAC Topology-Spatial Graph Dataset
tags:
- BIM
- HVAC
- graph-learning
- topology
- spatial-graph
- link-prediction
- operation-and-maintenance
size_categories:
- 100K<n<1M
---

# 面向BIM运维的暖通空调系统拓扑-空间图数据集

HVAC Topology-Spatial Graph Dataset for BIM-Based Operation and Maintenance

## Dataset Description

This dataset is an anonymized topology-spatial graph dataset for HVAC systems in BIM-based operation and maintenance. It contains public local graph fragments extracted from real BIM HVAC models. Each public graph file preserves component categories, MEP system labels, local spatial attributes, and explicit connection edges, while removing original element identifiers and building-level spatial relationships.

## Data Scale

- Public graph fragments: 406
- Nodes: 363683
- Edges: 359086
- Connected public files: 2
- Public files with multiple connected components: 404
- Version: 1.0.0

## Directory Structure

```text
nodes/
  1_nodes.json
  ...
edges/
  1_edges.json
  ...
public_file_index.csv
dataset_statistics.csv
schema.json
checksums.sha256
CITATION.cff
LICENSE
```

## Data Format

Each `nodes/{id}_nodes.json` file stores a list of HVAC component nodes. Each `edges/{id}_edges.json` file stores explicit connections among nodes in the corresponding node file.

Node fields:

- `id`: anonymized integer node ID, local to each public graph file and starting from `10000`.
- `category`: HVAC component category.
- `type`: component type retained after anonymization.
- `mepSystem`: MEP system label list.
- `position`: local anonymized XYZ coordinate, not an absolute building coordinate.
- `extend`: approximate spatial extent.
- `rotation`: component rotation attributes.

Edge fields:

- `source`: source node ID in the same public graph.
- `target`: target node ID in the same public graph.
- `connectType`: connection type.

## Statistics Fields

- `is_connected`: whether the public file is a single connected graph.
- `component_count`: number of connected components in the public file.
- `largest_component_node_count`: node count of the largest connected component.
- `largest_component_node_ratio`: largest component node count divided by total node count.
- `isolated_node_count`: number of degree-zero nodes.
- `avg_edge_length`, `max_edge_length`, `p95_edge_length`: edge length statistics in local coordinates.

Some public files contain multiple connected components. This is intentional and results from topology fragmentation, spatial balancing, and privacy-preserving release granularity control.

## Anonymization

The public release applies the following anonymization steps:

1. Retains only HVAC-related component categories.
2. Splits and balances original models into local topology-spatial fragments.
3. Converts each public graph to a local coordinate system.
4. Applies local rigid transformations and random public file re-indexing.
5. Re-indexes node IDs independently in each public graph from `10000`.
6. Removes original `elementId`, original file names, source-model grouping, and reversible coordinate transformation parameters.

The public release does not include any private audit mapping. Do not publish private audit files if you maintain them locally.

## Intended Uses

This dataset can support research on:

- HVAC graph learning.
- Link prediction and connection repair.
- Local topology completion.
- Spatial constraint learning.
- Engineering rule mining.
- BIM-to-digital-twin or BIM-to-knowledge-graph preprocessing.

## Limitations

- Coordinates are local anonymized coordinates and should not be interpreted as absolute building coordinates.
- Public graph files may contain multiple connected components.
- Source-model grouping information is intentionally removed from the public release.
- The dataset is intended for non-commercial research and educational use under CC-BY-NC-4.0.

## Citation

Please cite this dataset if you use it. A preliminary citation template is provided in `CITATION.cff`. After a DOI is assigned by Hugging Face, Zenodo, ScienceDB, or another repository, please update the citation metadata with the final DOI.

## License

This dataset is released under the Creative Commons Attribution-NonCommercial 4.0 International (CC-BY-NC-4.0). See `LICENSE` for details.