Automating clash relevance filtering in BIM-based multidisciplinary coordination using machine learning

In a context where Machine Learning (ML) is reshaping the construction industry and where normative frameworks such as ISO 19650 govern BIM data management, this paper aims to automate the filtering of true and false clashes in 3D models coordination process, using machine learning (ML). A metadata extraction plug-in is developed to gather the necessary data for training ML models. Tests are conducted on BIM models to evaluate the plug-in's ability to identify and classify clashes, followed by a reimplementation of the solution within an existing BIM software environment. Validation, carried out through both technical testing and feedback from industry professionals, demonstrates the plug-in's functionality and its ability to replicate the decision-making process of a BIM coordinator in clash filtering. Intended for construction professionals this paper highlights the potential of AI to enhance BIM quality control while complying with regulatory standards and meeting the practical needs of the industry. • Metadata extraction plug-in for training ML models. • Evaluation on BIM models for clash identification and classification. • Implementation within existing BIM software environment. • Validation through technical testing and industry feedback. • Demonstrated functionality replicating BIM coordinator decision-making in clash filtering.