From 2D Drawings to BIM Models: An LLM-Assisted Approach

Building Information Modeling (BIM) has become a central platform for digital building representation; however, the creation and manipulation of BIM models remain largely manual, and accessible primarily to domain experts.At the same time, architectural information is still commonly produced in 2D drawing formats, creating inefficiencies in translating design intent into semantically rich 3D BIM models.Recent advances in large language models (LLMs) offer new possibilities for natural language-based interaction with complex digital systems, yet their integration with BIM for automated and scalable building modeling is still limited.This research addresses this gap by proposing an automated framework that integrates LLMs with BIM systems to support the transformation of 2D architectural drawings into parametric model generation and manipulation.The framework's capability is demonstrated by using a single-floor architectural plan as input; the system enables the scalable generation of multi-storey buildings.The results indicate that LLM-assisted BIM workflows can significantly reduce manual modeling effort and improve efficiency, while remaining computationally tractable for generating multi-storey buildings.This study demonstrates the potential of LLM-BIM integration as a foundation for more intuitive, efficient, and scalable building modeling workflows.