Scholarlink

<p>This study presents a comprehensive framework for predicting the shear strength of reinforced concrete (RC) deep beams using a data-driven and mechanism-based model. Employing a dataset sourced from extensive experimental studies of 587 RC deep beams, six machine learning (ML) models were assessed using various evaluation metrics. The Stacking ensemble model exhibited superior performance, yielding the highest coefficient of determination (<i>R</i><sup>2</sup>) value of 0.996 along with a low Root Mean Square Error (<i>RMSE</i>) of 9.248, Mean Absolute Error (<i>MAE</i>) of 7.006, Median Absolute Error (<i>MdAE</i>) of 6.339, and Mean Squared Logarithmic Error (<i>MSLE</i>) of 0.003. Comparative evaluations against five conventional mechanism models, such as ACI 318-18 and EC2 1-2004, highlighted the Stacking model's accuracy and stability. Feature importance and dependency analyses identified geometric parameter (<i>a</i>/<i>d</i>), concrete strength (<InlineEquation ID="IEq1"> <EquationSource Format="TEX">\(f_{c}{\prime}\)</EquationSource> <EquationSource Format="MATHML"><math> <mrow> <msub> <mi>f</mi> <mi>c</mi> </msub> <mo>′</mo> </mrow> </math></EquationSource> </InlineEquation>), and longitudinal reinforcement properties (<i>f</i><sub><i>y</i></sub>, <i>ρ</i><sub><i>l</i></sub>) as primary contributors to shear strength predictions. The findings offer engineers nuanced insights for optimizing RC deep beam designs.</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

Enhancing Predictive Accuracy in Shear Strength of RC Deep Beams: A Comprehensive Analysis Using Ensemble Machine Learning Models

  • 期刊论文

  • 发表日期：

    2025年10月24日

查看全文
  • Arslan Qayyum Khan,
  • Muhammad Huzaifa Naveed,
  • Muhammad Dawood Rasheed,
  • Amorn Pimanmas

摘要

This study presents a comprehensive framework for predicting the shear strength of reinforced concrete (RC) deep beams using a data-driven and mechanism-based model. Employing a dataset sourced from extensive experimental studies of 587 RC deep beams, six machine learning (ML) models were assessed using various evaluation metrics. The Stacking ensemble model exhibited superior performance, yielding the highest coefficient of determination (R2) value of 0.996 along with a low Root Mean Square Error (RMSE) of 9.248, Mean Absolute Error (MAE) of 7.006, Median Absolute Error (MdAE) of 6.339, and Mean Squared Logarithmic Error (MSLE) of 0.003. Comparative evaluations against five conventional mechanism models, such as ACI 318-18 and EC2 1-2004, highlighted the Stacking model's accuracy and stability. Feature importance and dependency analyses identified geometric parameter (a/d), concrete strength ( \(f_{c}{\prime}\) f c ′ ), and longitudinal reinforcement properties (fy, ρl) as primary contributors to shear strength predictions. The findings offer engineers nuanced insights for optimizing RC deep beam designs.