Scholarlink

<p>A number of machine learning (ML) algorithms using tree-based methodologies were developed to forecast the ultimate strain (ε<sub>cu</sub>) of circular columns wrapped in aramid fiber-reinforced polymer (AFRP). This effort aimed to develop and evaluate ML techniques for the assessment of ε<sub>cu</sub> via the use of a dataset including 156 experimental results from 19 published research papers. This purpose led to the development of the decision tree (DT). The hyperparameters of DT, as established by the Beluga whale optimization algorithm (BWOA) and the Golden jackal optimization algorithm (GJOA) (DT(B) and DT(G)), greatly influence DT’s effectiveness. The practical novelty of this research lies in its development of more accurate and reliable predictive models for the <InlineEquation ID="IEq1"> <InlineMediaObject> <ImageObject Color="BlackWhite" FileRef="41939_2025_796_Article_IEq1.gif" Format="GIF" Height="12" Rendition="HTML" Resolution="72" Type="Linedraw" Width="23" /> </InlineMediaObject> <EquationSource Format="TEX">\({\varepsilon }_{cu}\)</EquationSource> <EquationSource Format="MATHML"><math> <msub> <mi>ε</mi> <mrow> <mi mathvariant="italic">cu</mi> </mrow> </msub> </math></EquationSource> </InlineEquation> of circular columns wrapped in AFRP by leveraging advanced optimization algorithms. These innovations enhance the precision of structural design and safety assessments, providing engineers with a robust tool for optimizing material selection, reducing costs, and improving the safety and performance of AFRP-wrapped structures across various conditions. According to forecast reliability and measurement variability, U95 evaluates uncertainty accurately. This shows a measure of dependability for informed decision-making. In learning and assessment, DT(G) had the lowest U95 index values—0.2893 and 0.2261. These values were below DT(B)’s 0.323 and 0.2476 throughout training and assessment. Based on the variance percentage that was used, the variation percentage among the two models for these metrics—which is at least 8% and sometimes 42%—demonstrates the predictive power and dependability of the DT(G).</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

Optimized decision tree algorithms to estimate ultimate strain of concrete wrapped by aramid fiber-reinforced polymer

  • 期刊论文

  • 发表日期：

    2025年3月1日

查看全文
  • Yangyang Guo

摘要

A number of machine learning (ML) algorithms using tree-based methodologies were developed to forecast the ultimate strain (εcu) of circular columns wrapped in aramid fiber-reinforced polymer (AFRP). This effort aimed to develop and evaluate ML techniques for the assessment of εcu via the use of a dataset including 156 experimental results from 19 published research papers. This purpose led to the development of the decision tree (DT). The hyperparameters of DT, as established by the Beluga whale optimization algorithm (BWOA) and the Golden jackal optimization algorithm (GJOA) (DT(B) and DT(G)), greatly influence DT’s effectiveness. The practical novelty of this research lies in its development of more accurate and reliable predictive models for the \({\varepsilon }_{cu}\) ε cu of circular columns wrapped in AFRP by leveraging advanced optimization algorithms. These innovations enhance the precision of structural design and safety assessments, providing engineers with a robust tool for optimizing material selection, reducing costs, and improving the safety and performance of AFRP-wrapped structures across various conditions. According to forecast reliability and measurement variability, U95 evaluates uncertainty accurately. This shows a measure of dependability for informed decision-making. In learning and assessment, DT(G) had the lowest U95 index values—0.2893 and 0.2261. These values were below DT(B)’s 0.323 and 0.2476 throughout training and assessment. Based on the variance percentage that was used, the variation percentage among the two models for these metrics—which is at least 8% and sometimes 42%—demonstrates the predictive power and dependability of the DT(G).