Machine learning models for seismic analysis of buckling-restrained braced frames

The current research addresses the need for efficient analysis, design, and post-earthquake damage assessment of buckling-restrained braced frames by developing machine learning (ML) models to predict multiple engineering demand parameters (EDPs) under various seismic loading conditions. To this end, a database of 16,694 records is formulated using numerical simulations. Then, nine supervised ML algorithms are optimised through hyperparameter tuning and validated to identify the most effective prediction model. The extreme gradient boosting (XGBoost) model demonstrated superior performance in estimating the inter-storey drift ratio (IDR), residual drift ratio (RDR), maximum ductility demand (mu max), and cumulative ductility demand (mu cum). Consequently, a user-friendly graphical user interface is devised for its seamless implementation. Finally, interpretable ML techniques, such as Shapley additive explanations (SHAP) and accumulated local effects, are applied to the XGBoost model to discern key input parameters and prediction trends. Pseudo-spectral acceleration at 2.0 s is identified as the most influential variable for predicting IDR, RDR, and mu max, while Arias intensity is the most significant for predicting mu cum. The top-ranking earthquake parameters identified through SHAP aid structural designers in assessing optimal intensity measures for fragility analysis.