1mbermek@gatech.edu 2,3{dennis.shelden|russell.gentry}@design.gatech. edu Mass Engineered Timber products provide a unique opportunity in configuring panelized building systems that are suitable for both prefabrication and onsite customization. The structural nature of these large section elements also brings about the need for a coordinated design-fabrication-assembly workflow. These products can assume different geometric configurations and their behaviour can be approximated globally by simplifying framing schemas. Current BIM Interoperability standards such as STEP or IFC already acknowledge and support the interconnected nature of component properties, yet these Data Models are component focused. Expanding on the relationships between components and using sets to define part to whole, or exteriority relationships could yield a more flexible and agile querying of building information.This would be a framework fit for automated feature derivation and rule based design applications.
| Authors | Mehmet Sinan Bermek and Dennis Shelden and T. Russel Gentry |
|---|---|
| Year | 2019 |
| Venue | Proceedings of the 37th International Conference on Education and Research in Computer Aided Architectural Design in Europe (eCAADe) \& 23rd Conference of the Iberoamerican Society Digital Graphics [Volume 2] |
| DOI | 10.52842/CONF.ECAADE.2019.2.789 |
| Source Database | Seed (CumInCAD) |
| Bridge-to-GNN | Category Seed |
| GNN Architecture | Not yet extracted from PDF |
| Graph Encoding | Structural Connectivity Graph (inferred from title) |
| AEC Task | Not yet extracted from PDF |
| Cohort | Early Infrastructure (2020–2021) |
| Dataset Size | 795 models (heuristic — verify) |
|---|---|
| Implementation Framework | Neo4j (heuristic — verify) |
| Primary Metric | Recall (heuristic — verify) |
| Quality Assessment | 2/6 · rigor: Low · Code/Data: ✓Baseline: ✓ (heuristic — verify) |
| GNN architectures detected | Knowledge Graph Embedding (TransE/RotatE/DistMult) |
|---|---|
| Frameworks / libraries | Neo4j / Cypher |
| Dataset stats found | 795 models |
| Quality (heuristic, 6-flag) | 0/6 · rigor: Low Code/Data: —Benchmark: —Baseline: —CV/Split: —Ablation: —Reproducible: — |
Part of the GML/GNN in AEC Systematic Review (PRISMA 2020) — 112 papers, 2020–2026