Scholarlink

<p>Reinforced concrete deep beams (RCDBs) exhibit complex nonlinear shear behaviour, making accurate strength prediction challenging for traditional models. This study presents a data-driven approach that combines k-fold cross-validated stepwise regression with a graphical user interface (GUI) for predicting RCDB shear strength. Using a dataset of 789 experimental cases, both linear and polynomial stepwise models were developed. The polynomial model (SPR) outperformed prior models, achieving an R<sup>2</sup> of 0.964, with effective depth and beam width identified as key influencing factors. Comparative analysis using Taylor diagrams, violin plots, and error thresholds confirmed SPR’s superior predictive accuracy, with 100% of predictions falling within 13% error. The interactive GUI enables users to adjust input parameters and visualize results, bridging analytical rigor with engineering usability. Finally, the proposed model provides a practical and accurate tool for RCDB shear prediction.</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

Deep beam shear prediction via K-fold cross-validated stepwise regression and a graphical user interface: a comparative analysis with state-of-the-art models

  • 期刊论文

  • 发表日期：

    2025年8月12日

查看全文
  • Maher K. Abbas,
  • Iman Kattoof Harith

摘要

Reinforced concrete deep beams (RCDBs) exhibit complex nonlinear shear behaviour, making accurate strength prediction challenging for traditional models. This study presents a data-driven approach that combines k-fold cross-validated stepwise regression with a graphical user interface (GUI) for predicting RCDB shear strength. Using a dataset of 789 experimental cases, both linear and polynomial stepwise models were developed. The polynomial model (SPR) outperformed prior models, achieving an R2 of 0.964, with effective depth and beam width identified as key influencing factors. Comparative analysis using Taylor diagrams, violin plots, and error thresholds confirmed SPR’s superior predictive accuracy, with 100% of predictions falling within 13% error. The interactive GUI enables users to adjust input parameters and visualize results, bridging analytical rigor with engineering usability. Finally, the proposed model provides a practical and accurate tool for RCDB shear prediction.