Equation discovery for post-earthquake roof drift ratio estimation using sparse regression and genetic algorithm
Journal
Journal of Building Engineering
Journal Volume
128
Start Page
116503
ISSN
23527102
Date Issued
2026-06-15
Author(s)
Abstract
Rapid estimation of post-earthquake deformation demand is critical for structural assessment, since drift is a key indicator of damage to structural and non-structural components. This study develops a new equation for roof drift ratio (RDR) estimation based on a genetic algorithm and sparse regression using seismic physics parameters. The genetic algorithm produces candidate functions through an evolutionary process, while sparse and ridge regression reduce equation complexity and form candidate equations. Physics-based parameters such as structural period, Peak Ground Velocity (PGV), Peak Ground Acceleration (PGA), and spectral acceleration (Sa) are considered. The equation discovery process involves using data collected from 46 reinforced concrete structures across 16 different sources, totaling 163 data samples. The accuracy and stability of the proposed RDR estimation equation are evaluated using the mean absolute error (MAE), mean absolute percentage error (MAPE), root mean square error (RMSE), and standard deviation. To further assess the performance, a neural network (NN) model is also developed and used as a baseline. An additional independent dataset is further used for external evaluation to examine the applicability of the proposed equation under different dataset characteristics. Results indicate that the proposed approach achieves a 15% to 30% reduction in RMSE compared to existing methods. Furthermore, the proposed equation shows favorable performance across the datasets considered in this study, with more than 96% predictions falling within absolute RDR differences of 0.5% and 1.0% from the measured values. These findings demonstrate the potential of the proposed equation as a rapid estimator of post-earthquake roof-level deformation demand.
Subjects
Equation discovery
Genetic algorithm
Post-earthquake assessments
Roof drift ratio estimation
Sparse regression
Publisher
Elsevier Ltd
Type
journal article