As construction projects become more complex, effective model coordination and clash resolution processes are critical for project success. Whereas clash detection is highly automated in existing software to identify conflicts between disciplines in building information models (BIM), review and resolution of these clashes remain primarily manual and, hence, labor-intensive. Despite previous research exploring various technologies and artificial intelligence to support this process, gaps remain. Specifically, no study aims to streamline the clash analysis and support resolution process in a manner that aligns with how BIM experts operate in real-world projects. In this study, a three-step framework is proposed to streamline BIM coordination through enhanced clash analysis and resolution. Multiple predictive machine learning algorithms were applied to validate the framework using real-world clash data. For clash relevance prediction, models such as artificial neural network (ANN) and support vector machine (SVM) achieved more than 80% accuracy and weighted F1 score, with precision up to 88%. For clash risk level prediction, random forest and gradient boosting reached up to 70% accuracy, whereas ANN offered the best balance between precision and recall. This study advances the field of BIM coordination and clash management by introducing a novel framework that integrates with real-world design processes, and by demonstrating how machine learning algorithms can achieve high accuracy in predicting clash relevance and risk level. Additionally, this study provides practical contributions to industry practitioners, by demonstrating a methodology to contextualize clash data and streamline model coordination by filtering out unnecessary work.
Koo et al. (Mon,) studied this question.