Development of a prefabricated construction productivity estimation model through BIM and data augmentation processes

The accuracy of productivity estimates remains a significant challenge due to limited data availability. This research addresses the need for precise productivity estimation in construction by integrating data augmentation techniques, onsite time study data, and Building Information Modeling (BIM) for automated quantity take-offs and design complexity analysis of steel connections. By examining design complexity, the method provides productivity estimates for project zones, sequences, and individual components, improving overall production management. Four data augmentation techniques—normal noise, interpolation, clustering, and Bayesian Linear Regression—were evaluated to enhance time study data. The augmented dataset was used to train an Artificial Neural Network, validated through case studies. The study identified the normal noise method as the most effective, significantly improving time estimation accuracy. Specifically, the proposed approach yielded a 58%–71% enhancement over current industry estimates and a 2.1%–31.1% improvement compared to models without data augmentation. This research enables managers to optimize resource allocation and reduce potential project delays.