Background/Motivation

The digital transformation in the construction industry is progressing – but one key area of knowledge remains largely untapped: the structural 2D execution details that are used extensively in the planning and execution phases.

These details are usually available as CAD files or PDF drawings, are not semantically structured and can only be searched, analyzed or reused manually. As a result, valuable experience from previous projects is lost, which could be used for efficient and sustainable planning.

Contents and goals

This is where LiveDetail comes in: the aim of the exploratory project is to investigate the viability of an AI-supported workflow for the automated recognition, structuring and reuse of technical 2D planning details.

The focus is on developing a prototypical approach that will enable planning offices to access existing detailed knowledge in an intelligent way in the future – supported by artificial intelligence, semantic data models and ecological evaluation criteria.

The planned process combines methods of machine image analysis (vision transformers, convolutional neural networks), semantic classification (graph neural networks) and natural language-based interaction (large language models).

Methodical Approach

The aim is to recognize and classify component connections, layer structures and materials from existing detailed plans and convert them into a structured, openBIM-like or compatible (ifc) format. At the same time, simplified life cycle assessment parameters with a focus on recyclability are recorded in order to enable the ecological impact of individual detail variants to be assessed at an early stage of planning.

An interaction concept tests the intuitive querying of digitized details using AI language models – for example, via a voice-based search for 'facade connections in solid wood construction'. The aim is to enable planners and architects to integrate high-quality and sustainable detail variants into their planning in a time-saving manner.

Expected Results

The innovative nature of the project lies in the linking of geometric, semantic and ecological information within a continuous digital workflow – a standard that has not yet been achieved in everyday planning.

The project addresses key challenges of climate-neutral urban development while creating new perspectives for digitalization, knowledge management and the circular economy in the construction industry.

The one-year exploratory phase will be used to explore the methodological foundations, technical potential and organizational conditions for a follow-up project.

Project management

Digital Findet Stadt GmbH

Project or cooperation partners

• AIT Austrian Institute of Technology GmbH
• Delta Pods Architects ZT GmbH
• FH Salzburg
• RWT PLUS ZT GmbH

Digital Findet Stadt GmbH
Prinz-Eugen-Straße 18/1/7
A-1040 Vienna
Tel.: +43 (664) 418 92 14
E-mail: office@digitalfindetstadt.at
Web: www.digitalfindetstadt.at