Scholarlink

<p>Fibre reinforcement in GPC significantly contributes to its ability to resist deformation and inhibit the propagation of cracks during tensile and flexural loading conditions. However, how the compressive strengths of fibre-reinforced geopolymer concrete (FRGPC), with steel hook fibres, can be predicted has been difficult due to the intricate relationship of the various components of the concrete mix that require adequate consideration during the process. Although artificial intelligence/machine learning (ML) has been indicated by various researchers as an alternative in modelling geopolymer concrete data, its use in FRGPC data has been on the periphery of research-based data due to the intricate mechanisms of concrete formation. In this study, an attempt has been made to produce an optimal data set using the Taguchi orthogonal array. In this regard, the proposed framework has produced a data set with 110 data points, covering all the necessary input parameters consisting of 16 input variables with varying percentages of GGBS and fly ash in the concrete mix. Considering the above results obtained with the Taguchi approach, an implementation of Linear Regression (LR), Decision Tree (DT), and Random Forest (RF) models was carried out with regards to compressive strength predictions. To evaluate the accuracy of the models, a set of statistical measures were used, including mean absolute error, mean squared error, root mean square error, mean absolute percentage error, and coefficient of determination, with k-fold cross-validation used in the models to avoid overfitting. Based on the results, a high predictive efficiency was recorded regarding all models used, while the Random Forest algorithm recorded higher accuracy with an R² of 0.906 and an RMSE of 3.822. The most important variables that affect compressive strength were identified in the sensitivity analysis as age of the specimens, alkaline solution to GGBS mass ratio, and dosage of sodium hydroxide. Equally important was the finding that Taguchi and ML were an efficient framework in the prediction of FRGPC compressive strengths.</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

Predictive modelling of fibre-reinforced geopolymer concrete strength using ensemble learning techniques

  • 期刊论文

  • 发表日期：

    2026年5月21日

查看全文
  • Debidatta Behuria,
  • Debasree,
  • Nihar Ranjan Mohanta,
  • Tapas Mohanty,
  • Trilochan Sahu

摘要

Fibre reinforcement in GPC significantly contributes to its ability to resist deformation and inhibit the propagation of cracks during tensile and flexural loading conditions. However, how the compressive strengths of fibre-reinforced geopolymer concrete (FRGPC), with steel hook fibres, can be predicted has been difficult due to the intricate relationship of the various components of the concrete mix that require adequate consideration during the process. Although artificial intelligence/machine learning (ML) has been indicated by various researchers as an alternative in modelling geopolymer concrete data, its use in FRGPC data has been on the periphery of research-based data due to the intricate mechanisms of concrete formation. In this study, an attempt has been made to produce an optimal data set using the Taguchi orthogonal array. In this regard, the proposed framework has produced a data set with 110 data points, covering all the necessary input parameters consisting of 16 input variables with varying percentages of GGBS and fly ash in the concrete mix. Considering the above results obtained with the Taguchi approach, an implementation of Linear Regression (LR), Decision Tree (DT), and Random Forest (RF) models was carried out with regards to compressive strength predictions. To evaluate the accuracy of the models, a set of statistical measures were used, including mean absolute error, mean squared error, root mean square error, mean absolute percentage error, and coefficient of determination, with k-fold cross-validation used in the models to avoid overfitting. Based on the results, a high predictive efficiency was recorded regarding all models used, while the Random Forest algorithm recorded higher accuracy with an R² of 0.906 and an RMSE of 3.822. The most important variables that affect compressive strength were identified in the sensitivity analysis as age of the specimens, alkaline solution to GGBS mass ratio, and dosage of sodium hydroxide. Equally important was the finding that Taguchi and ML were an efficient framework in the prediction of FRGPC compressive strengths.