Machine learning improves missing connections in building equipment graphs
Automated Brick Schema Knowledge Graph Completion Across Buildings
Machine Learning
Summary
Buildings have many sensors and equipment connected in complex ways, and describing these connections is important for smart building controls. The Brick schema is a way to map these relationships, but building these maps by hand is slow and hard. The authors test computer methods to guess the missing connections automatically, comparing three techniques on different building datasets. They find that one method, CompGCN, works best on some tests, but no single method is perfect across all cases.
What this means in practice
- •For building automation teams: Automatically fill in missing equipment and sensor connections in building models to speed up setup and reduce manual work.
- •For software developers for smart buildings: Incorporate AI-driven knowledge graph completion to improve general control applications that rely on complete building metadata.
- •For facility management software vendors: Create enhanced software products that offer automated completion of building metadata graphs to reduce on-site survey effort.$Commercial implications: This paper enables software products that auto-complete building equipment metadata graphs, reducing manual input needed by facility managers.
Authors
Daniel Ma, Jalil Huseynov, Xinyi Pan, Shaked Shalev, Zhang Chi, Ai Yamamoto, Sohei Arisaka
Abstract
The Brick schema represents building equipment, sensors, spaces, and the relationships between them as typed, multi-relational graphs, providing machine-readable metadata that supports generalizable control applications. Current methods for producing complete Brick models require extensive manual examination of building schematics and point lists. We observe that as Brick models are knowledge graphs, inferring missing relations is naturally posed as a knowledge graph completion problem. To the best of our knowledge, this is the first application of knowledge graph completion to Brick model generation. We compare three methods, gradient-boosted trees, Pairwise Brick Relation Ranking, and CompGCN, on the relations feeds, hasPart, and hasPoint, which account for the majority of the schematic-reading work in building a Brick model by hand. Models are evaluated on held-out MORTAR buildings and out-of-distribution Kaggle and Building Time Series data. Although CompGCN is the strongest performer on held-out buildings under metrics resistant to class imbalance, leading macro-F1 and AUPRC, no method generalizes uniformly across datasets or relations.