Scholarlink

<p>This study investigates the seismic performance of 15-storey reinforced concrete (RC) frames with slope variations of 0°, 10°, and 20°, incorporating soft stories at 2nd ,4th ,6th and 3rd ,6th ,9th floors. Situated on sloping ground, these soft-story configurations significantly increase seismic vulnerability. Dynamic analyses using Response Spectrum Analysis (RSA) and Time History Analysis (THA) were conducted in ETABS, with the 1966 Peru Lima Earthquake (Mw 8.1, depth 38&#xa0;km) accelerogram as input. RSA was used to evaluate peak displacement, story shear, and drift, confirming codal compliance per IS 1893:2016, with maximum top-storey displacements of 33–34&#xa0;mm for 20° slope models, within the 0.004&#xa0;h drift limit. In contrast, THA revealed much larger time-dependent nonlinear displacements up to 75&#xa0;mm, 150&#xa0;mm, and 230&#xa0;mm at the 2nd, 4th, and 6th floors, and 100&#xa0;mm, 240&#xa0;mm, and 370&#xa0;mm at the 3rd, 6th, and 9th floors highlighting severe vulnerability at soft stories. Machine learning models (Decision Tree, Random Forest, XGBoost) were trained on 288 structural response samples to predict seismic behavior. These models achieved high accuracy with R² values of 0.91 (Decision Tree), 0.94 (Random Forest), and 0.96 (XGBoost), and low RMSE between 0.042 and 0.058. SHAP feature analysis indicated story level, drift, and shear as key predictors. The study shows that while RSA ensures codal compliance, THA captures localized collapse mechanisms in soft-story buildings on slopes. Combining RSA, THA, and machine learning provides a comprehensive framework for performance-based seismic assessment, improving resilience in irregular RC buildings located in high seismic risk areas.</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 vulnerability assessment of soft-story RC buildings on inclined terrain using machine learning models

  • 期刊论文

  • 发表日期：

    2025年11月10日

查看全文
  • M. S. Ujwal,
  • G. Shiva Kumar,
  • M. G. Pranav.,
  • K. Sahana

摘要

This study investigates the seismic performance of 15-storey reinforced concrete (RC) frames with slope variations of 0°, 10°, and 20°, incorporating soft stories at 2nd ,4th ,6th and 3rd ,6th ,9th floors. Situated on sloping ground, these soft-story configurations significantly increase seismic vulnerability. Dynamic analyses using Response Spectrum Analysis (RSA) and Time History Analysis (THA) were conducted in ETABS, with the 1966 Peru Lima Earthquake (Mw 8.1, depth 38 km) accelerogram as input. RSA was used to evaluate peak displacement, story shear, and drift, confirming codal compliance per IS 1893:2016, with maximum top-storey displacements of 33–34 mm for 20° slope models, within the 0.004 h drift limit. In contrast, THA revealed much larger time-dependent nonlinear displacements up to 75 mm, 150 mm, and 230 mm at the 2nd, 4th, and 6th floors, and 100 mm, 240 mm, and 370 mm at the 3rd, 6th, and 9th floors highlighting severe vulnerability at soft stories. Machine learning models (Decision Tree, Random Forest, XGBoost) were trained on 288 structural response samples to predict seismic behavior. These models achieved high accuracy with R² values of 0.91 (Decision Tree), 0.94 (Random Forest), and 0.96 (XGBoost), and low RMSE between 0.042 and 0.058. SHAP feature analysis indicated story level, drift, and shear as key predictors. The study shows that while RSA ensures codal compliance, THA captures localized collapse mechanisms in soft-story buildings on slopes. Combining RSA, THA, and machine learning provides a comprehensive framework for performance-based seismic assessment, improving resilience in irregular RC buildings located in high seismic risk areas.