Knowledge Graphs for Multidisciplinary Co-Design: Introducing RDF to BHoM

Diellza Elshani, Alessio Lombardi, Al Fisher, Steffen Staab, Daniel Hernández, Thomas WortmannLDAC@ESWC 2022 · OpenReview

Introduction

In the Architecture, Engineering, and Construction (AEC) industry, a single building project is represented by a fragmented collection of digital models. Architects, structural engineers, and mechanical engineers often work in silos, leading to data inconsistencies that are frequently only discovered during construction. To address this, Elshani et al. (LDAC 2022) proposed a bridge between the Building Habitat object Model (BHoM) and Semantic Web standards (RDF/OWL), enabling the creation of multidisciplinary Knowledge Graphs for early-stage "Co-Design."

The Interoperability Challenge

While the Industry Foundation Classes (IFC) standard aims for software interoperability, it is often criticized as too monolithic for the iterative nature of early design. BHoM offers a more flexible, federated approach using "adapters" to translate data between tools like Revit and Rhino. However, BHoM lacks a formal mechanism to define relationships across different disciplinary views. Without these semantic links, automated cross-disciplinary checks—such as determining if an architectural change impacts structural integrity—remain difficult.

Methodology: Mapping BHoM to RDF

The researchers developed a pipeline to transform BHoM’s object-oriented structure into a Semantic Web format. This involved three primary components:

  1. Mapping Logic: A formal translation table was established between C# (BHoM’s native language) and RDF/OWL. For example, BHoM Classes map to owl:Class, and Interfaces map to rdfs:subClassOf.
  2. Semantic Enrichment: Since BHoM types often lack explicit relationship definitions, the authors introduced an IRelation interface and properties like hasProperty to serve as graph edges.
  3. The RDF_Toolkit: A custom C# tool was built to parse the BHoM codebase via reflection, automatically generating ontologies for over 1,200 classes.
Table 1: Mapping BHoM components to Semantic Web equivalents.

The primary dataset for this work was the BHoM codebase itself, supplemented by a "Building Demonstrator" test project consisting of core objects like rooms, walls, and columns to validate geometric dependencies.

Key Insights and Results

The study revealed that BHoM’s architecture is inherently well-suited for ontological conversion because it strictly separates data (object models) from functionality (engines). Furthermore, BHoM’s use of interfaces (e.g., IElement1D for linear elements) provides a natural taxonomic hierarchy for rdfs:subClassOf relationships.

The RDF_Toolkit successfully converted the extensive BHoM library into domain-specific ontologies (e.g., BH.oM.Structure). A small-scale building test generated a graph of 231 triples and 47 unique classes. The resulting complex relationships were visualized using WebVOWL, demonstrating a cohesive, navigable multidisciplinary graph.

Figure 1: Visualization of the multidisciplinary AEC data as a cohesive knowledge graph.

Discussion and Future Work

While this prototype proves the feasibility of the approach, several limitations remain. The current use of generic labels like HasProperty can lead to semantic ambiguity; future iterations will require more specific naming conventions. Additionally, while the tool effectively translates schemas, further work is needed to integrate "live" instance data from design tools like Grasshopper 3D. Finally, aligning BHoM-RDF with established industry ontologies, such as the Building Topology Ontology (BOT), is a priority for future development.

Summary updated

BibTeX
@inproceedings{elshani2022knowledge,
  author = {Diellza Elshani and Alessio Lombardi and Al Fisher and Steffen Staab and Daniel Hernández and Thomas Wortmann},
  title = {{Knowledge Graphs for Multidisciplinary Co-Design: Introducing RDF to BHoM}},
  booktitle = {LDAC@ESWC},
  year = {2022},
  url = {https://openreview.net/forum?id=AGmj5dvdwF},
}
Figure