BIM and Machine Learning Integration in The Design Phase, Using ML to Reduce the number of Irrelevant Clashes in Clash Detection Report

Integration of deep learning and BIM in BIM based design coordination

Abstract

As a digital transformation tool for Architecture, Engineering, Construction, and Operation (AECO), building information modeling (BIM) has revolutionized project execution and operation phases and dramatically impacted design coordination. Various professional teams usually design construction projects, resulting in design clashes. BIM-based design coordination can detect these clashes early on, saving time and resources compared to traditional methods. However, mainly the clash detection report from BIM software consists of many irrelevant clashes, which reduces the validity and reliability of the report and forces consulting companies to hire experts to achieve better and more accurate use of the clash detection report. This research aims to leverage the power of machine learning (ML) to enhance the automation of the process of clash detection and lessen the need for the use of experts in clash detection. For this purpose, a supervised ML algorithm has been developed to classify the clashes into relevant and irrelevant clashes. The “Methodology” of this research is based on producing the clash detection report from Navisworks (NW) software and developing a YOLO algorithm to classify the clashes and automatically recognize the relevant clashes. The analysis and validation of the algorithm’s accuracy in comparison to human-reasoning analysis will be discussed in the “validation and result” section, followed by limitations, the algorithm’s potential for extension, and suggestions for future studies in the “Conclusion” section.