Scholarlink

<p>The growing concerns about global warming and its adverse impacts on the environment, particularly with Portland cement production, developed the need to search for sustainable alternatives in the context of the construction industry. This study presents a transformative approach by utilizing alkali-activated concrete (ACC) that leverages industrial by-products presented in cement kiln dust (CKD), fly ash (FA), and silica fume (SF) to produce an environmentally friendly binder while maintaining enhanced mechanical properties. Furthermore, this study integrates experimental investigation with advanced artificial intelligence (AI) models to predict the compressive strength of ACC. The experimental part investigates the utilization of different binder ratios (0–100%), activator concentrations (6–14 mol/L), activator to binder ratio (0.35–0.55), and the ratio of SiO<sub>2</sub>/NaOH (1.5–3.5) on the compressive strength of alkali-activated concrete (ACC). Furthermore, this study investigates the potential of various AI models, such as Multiple Linear Regression (MLR), Multi-Layer Perceptron (MLP), Probabilistic Neural Network (PNN), and Support Vector Machine (SVM) in predicting the compressive strength of ACC. In this regard, binder ratios, activator concentrations, activator to binder ratio, and the ratio of SiO<sub>2</sub>/NaOH were utilized as input, while compressive strength was utilized as output. Moreover, the performance of the adopted models was assessed utilizing various statistical matrices and graphical appraisals. The results from the experimental part reveal that utilizing 25% CKD, 50% FA, and 25% FA combined with 10–14&#xa0;mol/L of activator concentrations, activator to binder ratio of 0.55, and SiO<sub>2</sub>/NaOH ratio up to 3.5 recorded a compressive strength exceeding 30 Mpa. Meanwhile, the results of the AI models reveal that the MLP tends to provide superior performance, recording minimal prediction deviation and high prediction accuracy (R<sup>2</sup> = 0.9), followed by the PNN model. Finally, the sensitivity analysis results showed that FA and SF content, as well as the average weight, have the highest impact on compressive strength.</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

Sustainable alkali-activated concrete modified with cement kiln dust and fly ash: an ai-based study

  • 期刊论文

  • 发表日期：

    2025年5月30日

查看全文
  • Saif Saad Mansoor,
  • Faidhalrahman Khaleel,
  • Haitham Abdulmohsin Afan,
  • Jumaa Mohammed Ahmad,
  • Alaa H. AbdUlameer

摘要

The growing concerns about global warming and its adverse impacts on the environment, particularly with Portland cement production, developed the need to search for sustainable alternatives in the context of the construction industry. This study presents a transformative approach by utilizing alkali-activated concrete (ACC) that leverages industrial by-products presented in cement kiln dust (CKD), fly ash (FA), and silica fume (SF) to produce an environmentally friendly binder while maintaining enhanced mechanical properties. Furthermore, this study integrates experimental investigation with advanced artificial intelligence (AI) models to predict the compressive strength of ACC. The experimental part investigates the utilization of different binder ratios (0–100%), activator concentrations (6–14 mol/L), activator to binder ratio (0.35–0.55), and the ratio of SiO2/NaOH (1.5–3.5) on the compressive strength of alkali-activated concrete (ACC). Furthermore, this study investigates the potential of various AI models, such as Multiple Linear Regression (MLR), Multi-Layer Perceptron (MLP), Probabilistic Neural Network (PNN), and Support Vector Machine (SVM) in predicting the compressive strength of ACC. In this regard, binder ratios, activator concentrations, activator to binder ratio, and the ratio of SiO2/NaOH were utilized as input, while compressive strength was utilized as output. Moreover, the performance of the adopted models was assessed utilizing various statistical matrices and graphical appraisals. The results from the experimental part reveal that utilizing 25% CKD, 50% FA, and 25% FA combined with 10–14 mol/L of activator concentrations, activator to binder ratio of 0.55, and SiO2/NaOH ratio up to 3.5 recorded a compressive strength exceeding 30 Mpa. Meanwhile, the results of the AI models reveal that the MLP tends to provide superior performance, recording minimal prediction deviation and high prediction accuracy (R2 = 0.9), followed by the PNN model. Finally, the sensitivity analysis results showed that FA and SF content, as well as the average weight, have the highest impact on compressive strength.