This is the authors' abstract. We don't add key points for this paper.
The fundamental period (T) is a key parameter governing the seismic response of reinforced concrete (RC) buildings, yet conventional code-based formulations often fail for hillside structures with pronounced irregularities. This study evaluates T in irregular RC moment-resisting buildings using numerical simulation, and machine learning techniques. Twelve 3D buildings models reflecting typical sloping terrain construction were analysed using the finite element method by varying height, plan geometry, and substructure depth. It was observed that the number of substructure storeys below ground level and plan geometry significantly affects T for hillside RC buildings. Therefore, plan and elevation area are more reliable predictor than building height alone This influence is captured in a total of 720 datasets obtained from numerical simulations which were used to predict fundamental period via Ensemble learning algorithms namely, Light Gradient Boosting Machine (LGBM) and Random Forest (RF). Building height, opening percentage of infill, area of elevation and plan were considered for development of empirical models utilizing nonlinear curve fitting and Multivariate Adaptive Regression Splines (MARS). Feature importance of the parameters were studied using SHAP analysis. NLR model exhibited the highest correlation coefficient (≈0.87) and MARS model achieved an R2 of 0.9135 and RMSE of 0.1065 in comparison to existing formulae. This can be attributed to the underestimation of the influence of irregularities existing formulae. By incorporating plan and elevation irregularities, the proposed models improve the reliability of fundamental period and offer a pathway toward resilience-oriented updates to current design guidelines.
Answers are grounded in this paper and cite their sources.
No takes yet. Share an insight, caveat, or question.
Sahal et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: