Damping ratio of reinforced concrete walls under limited damage from shaking table test and data‐driven modeling

Abstract

Accurate estimation of the damping ratio is critical for evaluating floor response spectra and equipment seismic demands in nuclear power plant structures. This study investigates the elastic damping ratio of reinforced concrete walls under limited damage states through shaking table tests and data‐driven modeling. Two wall specimens with aspect ratios of 1.0 and 2.0 were tested under various superimposed masses and earthquake inputs, and their damping ratios were precisely extracted using a state‐space system identification method. By integrating these experimental results with existing literature, a comprehensive database of 337 records with limited damage was constructed. While advanced tree‐based machine learning models were evaluated via 5‐fold cross‐validation, an explicit six‐term nonlinear regression model was ultimately developed to ensure structural interpretability and prevent overfitting to sample‐specific noise. Utilizing sequential feature selection, the proposed equation incorporates four core physical variables: axial force, cracking moment to maximum moment ratio, natural frequency, and aspect ratio. The explicit model achieved a highly stable predictive performance with an R ² of 0.786 and an RMSE of 0.017 for the entire dataset, maintaining a robust cross‐validation R ² of 0.766.