Knowledge Graphs in the Modern Building

Key Takeaways

• Open‑FDD builds open‑source, on‑premise building knowledge graph
• Uses RDF, Brick, Haystack standards for semantic interoperability
• AI tags BACnet points, cutting manual modeling time
• Keeps data local, eliminating vendor lock‑in risk
• Enables standardized retro‑commissioning and faster fault detection

Summary

The Open‑FDD project is developing a free, open‑source knowledge graph for smart buildings that lives on‑premise rather than inside a proprietary vendor platform. By leveraging RDF‑based standards such as Brick, ASHRAE 223P and Haystack, it models every BACnet point and equipment relationship in a semantic graph. AI‑assisted tagging dramatically reduces the manual effort required to map point names to ontology tags, while the resulting graph can power edge‑based fault detection and analytics. The approach promises vendor‑neutral data ownership, faster deployments, and more consistent retro‑commissioning outcomes.

Pulse Analysis

Smart‑building data models have long been a hidden asset locked inside vendor ecosystems. Standards like Brick, ASHRAE 223P, and Project Haystack provide a common RDF‑based language, but most owners rely on the vendor’s independent data layer to store and query that information. When a vendor departs, the model—and often the underlying data—vanishes, forcing costly rebuilds. Open‑FDD flips this paradigm by placing the knowledge graph on a local server managed by the building’s IT team, ensuring the data and its semantic relationships remain a permanent part of the facility’s digital infrastructure.

The technical core of Open‑FDD starts with a full BACnet network scan, converting raw point data into Turtle‑encoded RDF triples. Leveraging the bacpypes3 Python library, the system automatically captures device addresses, point identifiers, and topology. What used to require weeks of manual tagging is now accelerated by large language models that accurately map point names—such as “SA‑T”—to Brick tags and fault‑detection inputs. This AI‑driven pipeline not only slashes labor costs but also improves tagging consistency, laying a reliable foundation for edge‑based analytics and real‑time fault detection.

For operators and consultants, the implications are immediate. A standardized, on‑premise knowledge graph enables rapid, repeatable retro‑commissioning across multiple facilities, as illustrated by the Veterans Affairs hospital example where analysts could generate consistent performance charts with a few clicks. By decoupling analytics from any single cloud provider, building owners retain control while still allowing third‑party services to layer advanced insights on top. As Open‑FDD matures, its vendor‑neutral, AI‑enhanced approach could become the de‑facto baseline for energy‑efficient, data‑rich building management, driving industry‑wide adoption of structured, portable building intelligence.