Many organisations are being challenged by the speed and scale at which data is generated today. It is becoming progressively tricky to derive valuable insights, especially when the data sources are diverse and siloed. To gain a competitive advantage, organisations are starting to explore the next generation of scalable data management approaches.
A Knowledge Graph is one way to go. It allows an organisation to connect and show meaningful relationships between data regardless of type, format, size, or location.
The Architecture, Engineering and Construction (AEC) industry has been built to organise and manage data by department and type. A knowledge graph breaks these walls and connects the data, allowing stakeholders to answer specific business questions without copying the data from its original sources. Let’s take, for instance, a building;
- How much of the data and information used to manage a building is dispersed across its different systems, zones, components, documentation or databases?
- How many clicks and reports do facility managers have to navigate in the interface of a Building Automation System to find an answer to a single Facility Management problem or get relevant results for a specific search?
These are the types of puzzles that knowledge graphs are engineered to solve. Perhaps a good question to ask now is ‘how can a knowledge graph be used to represent a building’s information? Is it even sufficient? The first part of my PhD research answers this question within the context of Building Automation.
Let’s start with what everyone knows, a Building Infomation Model curated in Revit. To extract specific information from a BIM model, we can adopt a smart filter to get us only what we need and ignore the rest. The choice of smart filter depends on the intended use case. In this research, I adopted Construction-Operations Building Information Exchange (COBie) ;
‘a data model that allows us to filter a BIM model and extract only the information that is relevant for the management and operation of the building. ‘
COBie data can be imported directly into Computerised Maintenance Management Systems (CMMS) and asset management software. The accompanying PDFs, drawings, and BIM files are usually stored and accessed through a secure server somewhere at the FM office.
Data siloes again, do you see them?
The COBie data is in the CMMS, the BIM models are somewhere else, the CAD drawings are on their island, and the same applies to other PDF documentation. To make COBie data as useful as possible, it must be easy to query and interoperable with other BIM data. Lucky for us, knowledge graphs are best suited for this task.
Semantic Web Technologies provide us tools and standards for building knowledge graphs using formalised building blocks called ontologies which are usually domain-specific. My research builds on top of these technologies and RDFLib, a 3rd party python library, to create a customisable script that reads BIM data from a COBie file and builds a knowledge graph. This COBie-based knowledge graph offers integration channels for BIM models, PDFs and other CMMS documentation scattered across multiple repositories. Now, this information/data can be used holistically with no boundaries.
However, this is not the only power that knowledge graphs give us and Part 2 of this series will cover this within the context of Machine Learning.