Title:
Estimating the Drift Capacity of Reinforced Concrete Columns Using Machine Learning (Prepublished)
Author(s):
Liam Pledger, Santiago Pujol, and Reagan Chandramohan
Publication:
Structural Journal
Volume:
Issue:
Appears on pages(s):
Keywords:
columns; drift capacity; machine learning; reinforced concrete
DOI:
10.14359/51749374
Date:
12/8/2025
Abstract:
A machine learning (ML) model is developed using a gradient-boosted decision-tree algorithm to estimate the drift capacity of reinforced concrete (RC) columns. A reliable estimate of the drift capacity of a structure is critical to both its design and assessment. The drift capacity of a structure is also broadly interpreted as a measure of its seismic vulnerability. The estimated drift capacity from the ML model is compared against that of existing methods using test results from a dataset of 341 RC columns subjected to cyclic loading. The mean of the ratio of measured to estimated drift capacity for the developed ML model was 1.0 with a coefficient of variation (CV) of 0.31. In comparison, the regression equation currently adopted in New Zealand and the US to estimate the drift capacity of RC columns has a mean of 3.13 and a CV of 1.07. Other empirical methods assessed in this study also led to large scatter and no discernible correlation between estimated and measured drift capacity. The developed ML model provides more accurate results than existing methods and can estimate the drift capacity for a broad range of RC columns. The developed model is published under an open-source license and is freely available to practitioners and researchers.