Scholarlink

<p>The research investigates to optimise the parameters affecting the strength of the geopolymer composites developed using AI models. This paper evaluates seven AI models such as Auto Regression, Polynomial Regression, Multiple linear Regression, Random forest Regression, Logistic Regression, Decision Tree Regression, Gaussian Process Regression and Quantile Regression. The dataset utilized for the experimental work was incorporated from the experimental data which was published by the researchers earlier. The input variables with which the prediction was using the machine learning systems were the fly ash, Ground granulated blast furnace slag, coarse aggregate, fine aggregate, Na<sub>2</sub>SiO<sub>3</sub>, NaOH, Na<sub>2</sub>O(dry), SiO<sub>2</sub>(dry), Water, Concentration of NaOH, Additional water content, superplasticizer, total water content, Initial curing time, Initial curing temperature, Initial cutting rest time and final curing temperature for geopolymer concrete compressive strength prediction. For this study, input feature groups were developed, such as G1: Characteristics that have a substantial effect on the strength of geopolymer concrete, G2: Characteristics related to the composition of fly ash, G3: Ratios that different researchers use to forecast the strength of geopolymer concrete and G4: The top 13 features from groups 1, 2, and 3 for optimizing the parameters which have a significant effect on the prediction of the compressive strength. When the comprehensive ranking analysis was carried out across the multiple statical metrics, Gaussian process regression has produced the most accurate results with a R<sup>2</sup> of 0.9763, thus demonstrating the exceptional accuracy at a minimal error level, followed by polynomial regression and Randon Forest with R<sup>2</sup> values 0.9642 and 0.9484. The results indicate that the Gaussian process regression has outperformed all other prediction models in prediction the accurate values for the concrete compressive strength. This predictive approach for estimating the compressive strength of alkali-activated composites significantly contributes to meeting cost efficiency and time-saving requirements.</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

Data-driven ensemble machine learning framework for reliable compressive strength prediction of alkali-activated composites

  • 期刊论文

  • 发表日期：

    2026年3月5日

查看全文
  • K. Ramujee,
  • D. Praseeda

摘要

The research investigates to optimise the parameters affecting the strength of the geopolymer composites developed using AI models. This paper evaluates seven AI models such as Auto Regression, Polynomial Regression, Multiple linear Regression, Random forest Regression, Logistic Regression, Decision Tree Regression, Gaussian Process Regression and Quantile Regression. The dataset utilized for the experimental work was incorporated from the experimental data which was published by the researchers earlier. The input variables with which the prediction was using the machine learning systems were the fly ash, Ground granulated blast furnace slag, coarse aggregate, fine aggregate, Na2SiO3, NaOH, Na2O(dry), SiO2(dry), Water, Concentration of NaOH, Additional water content, superplasticizer, total water content, Initial curing time, Initial curing temperature, Initial cutting rest time and final curing temperature for geopolymer concrete compressive strength prediction. For this study, input feature groups were developed, such as G1: Characteristics that have a substantial effect on the strength of geopolymer concrete, G2: Characteristics related to the composition of fly ash, G3: Ratios that different researchers use to forecast the strength of geopolymer concrete and G4: The top 13 features from groups 1, 2, and 3 for optimizing the parameters which have a significant effect on the prediction of the compressive strength. When the comprehensive ranking analysis was carried out across the multiple statical metrics, Gaussian process regression has produced the most accurate results with a R2 of 0.9763, thus demonstrating the exceptional accuracy at a minimal error level, followed by polynomial regression and Randon Forest with R2 values 0.9642 and 0.9484. The results indicate that the Gaussian process regression has outperformed all other prediction models in prediction the accurate values for the concrete compressive strength. This predictive approach for estimating the compressive strength of alkali-activated composites significantly contributes to meeting cost efficiency and time-saving requirements.