Background/Motivation

The FacilityQ project addresses the challenges associated with transferring building data from the design and construction phases to the operations phase. At its core is an AI-powered workflow that integrates structured information from BIM models with operational documents and technical metadata.

Methodical Approach

The methodological implementation is divided into three key development areas.

A phase- and role-based data model is being developed that is based on established standards such as ÖNORM A 6241-2 and the Level of Information Need (LOIN). This model describes all operation-relevant information and links to documents and enables automated assignment to operating classes and cost groups via an AKS generator based on ÖNORM B 1801-6.

Secondly, FacilityQ relies on AI-supported data enrichment: With the help of transformer-based natural language processing (NLP) models such as BERT, technical documents are automatically classified and assigned in accordance with standards – regardless of format or designation.

In addition, rule-based heuristics are used, for example, for the robust assignment of text content to cost groups or for the recognition of procedural notes. Other AI modules analyze image material, such as type plates of HVAC components, to extract relevant operating parameters ("Smart Datasheet Capture").

These heterogeneous data are linked via semantic knowledge graphs that model relationships between BIM objects, documents, and properties and serve as the basis for automated data transfers.

Thirdly, an open, plugin-free microservice architecture is being developed that will enable all functions to be integrated into a common data environment (CDE). The transfer to CAFM systems is carried out via standardized interfaces such as IFC, BCF, and IDS. Explainable AI is used for quality assurance and traceability, for example through visualizations of decision influences (e.g., using SHAP or LIME), confidence scores, or source references to standards and training data.

Expected Results

The research design is iterative: the modules developed are continuously validated using real project data. An extensive database of over 30 million square meters of building space is available for training the AI models.

The goal is to reduce the time and cost of data transfer from construction to operation by at least 50% while significantly increasing data quality and interoperability.

Project management

AIT Austrian Institute of Technology GmbH

Project or cooperation partners

• Digital Findet Stadt GmbH
• FCP Fritsch, Chiari & Partner ZT GmbH
• immOH! Energie und Gebäudemanagement GmbH
• ISH-Solutions GmbH
• M.O.O.CON GmbH
• TendX GmbH

Gerhard Zucker
Giefinggasse 2
A-1210 Vienna
Tel.: +43 (664) 235 19 21
E-mail: gerhard.zucker@ait.ac.at
Web: www.ait.ac.at