Title |
A Comparative Study of Machine Learning model and BIM Based Methods for Rebar Quantification |
Authors |
이하늘(Lee, Ha-Neul) ; 윤석헌(Yun, Seok-Heon) |
DOI |
https://doi.org/10.5659/JAIK.2025.41.7.261 |
Keywords |
BIM; Machine Learning; Rebar Quantity |
Abstract |
Accurately calculating rebar quantities in the early stages of construction projects is important for cost estimation and resource management.
This study introduces two automated methods for rebar quantity calculation: one based on Building Information Modeling (BIM) and another
using machine learning. Both approaches rely on detailed design data. The BIM-based method restructured rebar quantity formulas to use only
information that can be extracted directly from BIM models. When tested on a case study, it achieved an average error rate of 2.015 percent
for columns and 4.925 percent for typical beams, showing reliable performance during the detailed design phase. The machine learning-based
method estimated rebar quantities using the rebar-to-concrete ratio. Approximately 100 samples, including concrete volume, rebar quantities,
and construction years, were used for training. The LeakyReLU activation function produced the most accurate results, with an error rate of
9.73 percent, which meets AACE standards for detailed estimates. Both methods showed potential for improving accuracy and enabling
automation. However, the study was limited by the size of the dataset and its focus on columns, beams, and slabs. Future research should
expand the dataset and apply the methods to a wider range of structural components. These findings provide a strong basis for early-stage
rebar quantity estimation and support more informed decision-making in construction planning. |