AI-based automation of repetitive design steps in BIM modeling

Researchers	Changyu Du
Funding	Nemetschek SE
Project runtime	01.04.2023.-31.03.2026

The broad adoption of BIM and model-based design in construction is slowed down by the fact that many users see the use of BIM modeling tools as an extra effort when compared with conventional drawing-based design. In fact, the modeling process is largely governed by repetitive tasks that have to performed manually and are thus laborious and time consuming.  Automated next-best action recommendation in a sequential, dynamic and interactive context has gained high significance in social and business decision-making. AI-based approaches have been the basis for a series of methodologies and approaches that predict the next most probable outcome given current status. Since BIM modelling often contains series of repetitive actions taken by the user, the development and deployment of decision making smart predictive systems could significantly reduce the time required for BIM modelling. The vision of the project is to relief the user from knowledge encoding by letting the software learn which modeling steps are typically performed in which order and with what parametrization. This is realized applying the latest AI methodology tto implement the concept of “learning-by-modeling”, which infers the necessary knowledge to predict the most suitable sequence of the next modeling operations, providing recommendations to the user.