Scholarlink

<p>The use of naturally available materials such as metakaolin (MK) can greatly reduce the utilization of emission intensive materials like cement in the construction sector. This would reduce the stress on depleting natural resources and foster a sustainable construction industry. However, the laboratory determination of 28&#xa0;day compressive strength (C-S) of MK-based mortar is associated with several time and resource constraints. Thus, this study was conducted to develop reliable empirical prediction models to assess CS of MK-based mortar from its mixture proportion using machine learning algorithms like gene expression programming (GEP), extreme gradient boosting (XGB), multi expression programming (MEP), bagging regressor (BR), and AdaBoost etc. A comprehensive dataset compiled from published literature having five input parameters including water-to-binder ratio, mortar age, and maximum aggregate diameter etc. was used for this purpose. The developed models were validated by means of error metrics, residual assessment, and external validation checks which revealed that XGB is the most accurate algorithm having testing <InlineEquation ID="IEq1"> <InlineMediaObject> <ImageObject Color="BlackWhite" FileRef="41598_2025_1327_Article_IEq1.gif" Format="GIF" Height="17" Rendition="HTML" Resolution="72" Type="Linedraw" Width="19" /> </InlineMediaObject> <EquationSource Format="TEX">\(\:{\text{R}}^{2}\)</EquationSource> </InlineEquation> of 0.998 followed by BR having <InlineEquation ID="IEq2"> <InlineMediaObject> <ImageObject Color="BlackWhite" FileRef="41598_2025_1327_Article_IEq1.gif" Format="GIF" Height="17" Rendition="HTML" Resolution="72" Type="Linedraw" Width="19" /> </InlineMediaObject> <EquationSource Format="TEX">\(\:{\text{R}}^{2}\)</EquationSource> </InlineEquation> values equal to 0.946 while MEP had the lowest testing <InlineEquation ID="IEq3"> <InlineMediaObject> <ImageObject Color="BlackWhite" FileRef="41598_2025_1327_Article_IEq1.gif" Format="GIF" Height="17" Rendition="HTML" Resolution="72" Type="Linedraw" Width="19" /> </InlineMediaObject> <EquationSource Format="TEX">\(\:{\text{R}}^{2}\)</EquationSource> </InlineEquation> of 0.893. However, MEP and GEP algorithms expressed their output in the form of empirical equations which other black-box algorithms couldn’t produce. Moreover, interpretable machine learning approaches including shapely additive explanatory analysis (SHAP), individual conditional expectation (ICE), and partial dependence plots (PDP) were conducted on the XGB model which highlighted that water-to-binder ratio and sample age are some of the most significant variables to predict the C-S of MK-based cement mortars. Finally, a graphical user interface (GUI) was made for implementation of findings of this study in the civil engineering industry.</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

Interpretable machine learning approaches to assess the compressive strength of metakaolin blended sustainable cement mortar

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2025年6月3日

查看全文
  • Naseer Muhammad Khan,
  • Liqiang Ma,
  • Waleed Bin Inqiad,
  • Muhammad Saud Khan,
  • Imtiaz Iqbal,
  • Muhammad Zaka Emad,
  • Saad S. Alarifi

摘要

The use of naturally available materials such as metakaolin (MK) can greatly reduce the utilization of emission intensive materials like cement in the construction sector. This would reduce the stress on depleting natural resources and foster a sustainable construction industry. However, the laboratory determination of 28 day compressive strength (C-S) of MK-based mortar is associated with several time and resource constraints. Thus, this study was conducted to develop reliable empirical prediction models to assess CS of MK-based mortar from its mixture proportion using machine learning algorithms like gene expression programming (GEP), extreme gradient boosting (XGB), multi expression programming (MEP), bagging regressor (BR), and AdaBoost etc. A comprehensive dataset compiled from published literature having five input parameters including water-to-binder ratio, mortar age, and maximum aggregate diameter etc. was used for this purpose. The developed models were validated by means of error metrics, residual assessment, and external validation checks which revealed that XGB is the most accurate algorithm having testing \(\:{\text{R}}^{2}\) of 0.998 followed by BR having \(\:{\text{R}}^{2}\) values equal to 0.946 while MEP had the lowest testing \(\:{\text{R}}^{2}\) of 0.893. However, MEP and GEP algorithms expressed their output in the form of empirical equations which other black-box algorithms couldn’t produce. Moreover, interpretable machine learning approaches including shapely additive explanatory analysis (SHAP), individual conditional expectation (ICE), and partial dependence plots (PDP) were conducted on the XGB model which highlighted that water-to-binder ratio and sample age are some of the most significant variables to predict the C-S of MK-based cement mortars. Finally, a graphical user interface (GUI) was made for implementation of findings of this study in the civil engineering industry.