Scholarlink

<p>Geopolymer is increasingly being recognized as a sustainable alternative to conventional cementitious materials for commercial applications, offering several advantages. However, the development of geopolymer faces challenges in obtaining consistent strength due to the inconsistency in the chemical composition of alumina silicate precursors. The main purpose of conducting this study is to analyse the effect of main oxide ratios like Si/Al, Al/Na, Si/Na, and Na/H<sub>2</sub>O affect the physical and mechanical properties of the geopolymer composites. Therefore, six combinations of Fly Ash (FA) based and two combinations of Ground granulated blast furnace slag (GGBS) based geopolymer mixes were prepared with 10, 12 and 14 molar Sodium silicate (NaOH) solution were prepared for this experiment. Further, the results obtained from the experimental work were utilized to develop a correlation between different oxide ratio and compressive strength of geopolymer concrete using different machine learning techniques to understand the behaviour by considering different oxide ratios on the properties of geopolymer concrete. The study utilizes a total of four different machine learning algorithms, such as Random Forest, Gradient Boosting (GBR), AdaBoost, and stacking, which have been used to forecast the hardened properties of geopolymer concrete. The prediction performance of all the models for hardened properties was compared using a testing dataset, and it was noticed that the stacking models exhibit more accurate prediction than other Algorithm models. Analysis of the study reveals that the stacking model performed well in compressive strength for 3, 7 and 28&#xa0;days with a high correlation coefficient of R<sup>2</sup> (0.97701, 0.9564, 0.9513). Additionally, the stacking model exhibits the lowest RMSE values of 1.3065, 1.8022, and 1.8727 for compressive strength, for 3, 7 and 28&#xa0;days, respectively. Furthermore, a SHAP dependency analysis was performed to understand the significance of each parameter. It is observed from this study that Na/Si, followed by setting time is the most crucial parameter in predicting both 3-day and 7-day strength 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

Experimental and Machine Learning-Based Analysis of Oxide Ratios in Alumina-Silicate Geopolymer Formation

  • 期刊论文

  • 发表日期：

    2025年5月8日

查看全文
  • Ashwin N. Raut,
  • Anant Lal Murmu,
  • Sanjog Chhetri Sapkota,
  • Sourav Das,
  • Prasenjit Saha

摘要

Geopolymer is increasingly being recognized as a sustainable alternative to conventional cementitious materials for commercial applications, offering several advantages. However, the development of geopolymer faces challenges in obtaining consistent strength due to the inconsistency in the chemical composition of alumina silicate precursors. The main purpose of conducting this study is to analyse the effect of main oxide ratios like Si/Al, Al/Na, Si/Na, and Na/H2O affect the physical and mechanical properties of the geopolymer composites. Therefore, six combinations of Fly Ash (FA) based and two combinations of Ground granulated blast furnace slag (GGBS) based geopolymer mixes were prepared with 10, 12 and 14 molar Sodium silicate (NaOH) solution were prepared for this experiment. Further, the results obtained from the experimental work were utilized to develop a correlation between different oxide ratio and compressive strength of geopolymer concrete using different machine learning techniques to understand the behaviour by considering different oxide ratios on the properties of geopolymer concrete. The study utilizes a total of four different machine learning algorithms, such as Random Forest, Gradient Boosting (GBR), AdaBoost, and stacking, which have been used to forecast the hardened properties of geopolymer concrete. The prediction performance of all the models for hardened properties was compared using a testing dataset, and it was noticed that the stacking models exhibit more accurate prediction than other Algorithm models. Analysis of the study reveals that the stacking model performed well in compressive strength for 3, 7 and 28 days with a high correlation coefficient of R2 (0.97701, 0.9564, 0.9513). Additionally, the stacking model exhibits the lowest RMSE values of 1.3065, 1.8022, and 1.8727 for compressive strength, for 3, 7 and 28 days, respectively. Furthermore, a SHAP dependency analysis was performed to understand the significance of each parameter. It is observed from this study that Na/Si, followed by setting time is the most crucial parameter in predicting both 3-day and 7-day strength prediction.