Knowledge graphs provide a flexible framework for semantically-enriched representation of complex structures. The nodes corresponding to domain entities - events, components, situations, agents, locations, time periods - are connected by named relations - such as participant-of, part-of, connected-to, agent-of, precedes. Knowledge graphs combined with efficient graph algorithms have been used to manage transportation, telecommunication and social networks. More generally, however, knowledge graphs can represent problems involving large sets of heterogeneous entities that have complex networks of mutual relations. In such semantically rich knowledge graphs it is important to represent the types of the entities together with the types of their relations, attributes and restrictions: this is done with ontologies. The schema of a knowledge graph is typically defined as a combination of several different ontologies. Since each ontology is represented as a knowledge graph itself, and the entities in the domain knowledge graphs are directly linked to the classes and properties in the ontologies, the overall representation framework is flexible and allows the use of advanced reasoning techniques. 

Successful knowledge graph implementations have been (1) service knowledge graphs such as Google Knowledge Graph that enables advanced search functionalies and the Product Graph of Amazon that supports its e-commerce offerings, and (2) enterprise knowledge graphs that systematize the representation of the key information of an enterprise. Enterprise knowledge graphs often require the definition of a specific enterprise ontology, and the tranformation and loading of data from existing sources, such as relational databases. Enterprise knowledge graphs can provide a flexible data-centric basis for enterprise information systems, and they are clearly relevant also for companies in the construction sector. However, knowledge graphs can also be used for asset management and project management.

Asset information management about real estate, buildings, and building units is related to building information models (BIM), digital twins (BIM dynamically updated based on IoT/BAS data) and digital building logbooks (an evolving repository of the key data of a building). The complex asset information content can be modeled and represented using different existing ontologies for BIM (ifcOWL or BOT), BAS (Bricks), IoT (SSN/SOSA, Saref), and emerging models for digital building logbooks. Significant research and development on ontologies has been done in the field of Linked Building Data. 

In the management of construction projects knowledge graphs more challenging to apply, since information is created in a decentralized manner by several participating organizations, potentially based on different ontologies. The relatively short time-span of a project does not allow for a specification of a project specific ontology nor related transformations. Rather, the participating companies need to share data based on standardized ontologies, perhaps utilizing existing mappings between them. Fortunately, there are  many proposals for ontologies to manage construction and renovation projects, such as Digital Construction Ontologies (DiCon).

This talk will discuss the utilization of knowledge graphs in construction: existing work, advantages, and open challenges. The research was funded by BIM4EEB project, grant number 820660, as part of the EU Horizon 2020 research and innovation program.  

