Scholarlink

<p>Vertical setback irregularities—characterized by abrupt changes in a building’s geometry along its height—pose significant challenges in seismic design due to uneven mass and stiffness distribution. These discontinuities can amplify torsional effects, concentrate seismic forces, and increase vulnerability to structural damage. This study presents a data-driven approach to forecast critical seismic response parameters—time period, inelastic roof displacement, and storey drift—for reinforced concrete (RC) buildings exhibiting vertical setbacks. A comprehensive dataset of 549 building configurations was generated using ETABS, incorporating key input variables such as plan aspect ratio, building height (H), building parameter (r), and irregularity index (β). To ensure predictive accuracy, both parametric models (Multiple Linear Regression, Ridge, Bayesian Ridge) and non-parametric models (Decision Tree, Random Forest, AdaBoost, XGBoost, Gradient Boosting) were employed. Model performance was evaluated using R², MAE, MSE, MAPE, RMSE, Pearson correlation, K-fold cross-validation &amp; hyperparameter optimization. The study employed SHapley Additive exPlanations (SHAP) plots as post-hoc interpretability tools to analyze the contributions of structural parameters. Results show that total building height remains the dominant factor contributing to the predicted seismic response parameters. Through hyperparameter optimization, the Gradient Boosting Regressor achieved the highest accuracy with an R² of 0.99347, outperforming other techniques. Finally, as a key innovation, an interactive HTML-based graphical user interface (GUI) was developed to enable engineers and researchers to rapidly estimate seismic responses for setback-irregular buildings, streamlining analysis and reducing computational overhead. This research highlights the effectiveness of machine learning in addressing complex irregularities and supports its integration into seismic design workflows.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Seismic response prediction of Rc buildings with vertical setback irregularities: a machine learning framework with interactive HTML-based deployment

  • 期刊论文

  • 发表日期：

    2025年11月17日

查看全文
  • Anurag Wahane,
  • Sri Ram Krishna Mishra,
  • Pradeep Kumar Ghosh

摘要

Vertical setback irregularities—characterized by abrupt changes in a building’s geometry along its height—pose significant challenges in seismic design due to uneven mass and stiffness distribution. These discontinuities can amplify torsional effects, concentrate seismic forces, and increase vulnerability to structural damage. This study presents a data-driven approach to forecast critical seismic response parameters—time period, inelastic roof displacement, and storey drift—for reinforced concrete (RC) buildings exhibiting vertical setbacks. A comprehensive dataset of 549 building configurations was generated using ETABS, incorporating key input variables such as plan aspect ratio, building height (H), building parameter (r), and irregularity index (β). To ensure predictive accuracy, both parametric models (Multiple Linear Regression, Ridge, Bayesian Ridge) and non-parametric models (Decision Tree, Random Forest, AdaBoost, XGBoost, Gradient Boosting) were employed. Model performance was evaluated using R², MAE, MSE, MAPE, RMSE, Pearson correlation, K-fold cross-validation & hyperparameter optimization. The study employed SHapley Additive exPlanations (SHAP) plots as post-hoc interpretability tools to analyze the contributions of structural parameters. Results show that total building height remains the dominant factor contributing to the predicted seismic response parameters. Through hyperparameter optimization, the Gradient Boosting Regressor achieved the highest accuracy with an R² of 0.99347, outperforming other techniques. Finally, as a key innovation, an interactive HTML-based graphical user interface (GUI) was developed to enable engineers and researchers to rapidly estimate seismic responses for setback-irregular buildings, streamlining analysis and reducing computational overhead. This research highlights the effectiveness of machine learning in addressing complex irregularities and supports its integration into seismic design workflows.