Scholarlink

<p>Eco-friendly alternatives to expensive synthetic fibers and high-resource materials in concrete have been developing rapidly over recent years. Due to the cost-effective and renewable nature of jute fibers, they have great potential as a source of low-carbon concrete; however, their effects on the mechanical characteristics of concrete are still poorly defined. The integrated framework combining ensemble and interpretable machine learning algorithms is developed to investigate and predict the compressive strength of Jute Fiber-Reinforced Concrete (JFRC), based on a large and highly harmonized database of literature-based experimental data. Four existing Machine Learning (ML) methods, namely the Random Forest-Gradient Boosting (RF-Gb) Ensemble, AdaBoost, CatBoost, and Relevance Vector Machine (RVM), were used. It was determined that CatBoost and the RF-Gb method performed best in terms of predicting experimental results (compressive strength), and that there was excellent correlation between the predicted and actual experimental results. The SHAP-based interpretability analysis also revealed that flexural strength and tensile strength at split-tensile testing were the most significant factors affecting compressive strength, while jute-fiber dosage, workability, and replacement of fine aggregate were secondary influencing factors. These findings show the feasibility of using jute fibers as a sustainable reinforcement in the production of low-carbon concrete and that the application of interpretable ML can provide a means of supporting informed design of mixes for optimized, low-carbon, natural fiber-concrete products. Overall, the present study provides a data-driven and scalable framework that supports the development of high-performance and environmentally responsible, fiber-reinforced composites.</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 jute fiber-reinforced concrete: an integrated ensemble and interpretable machine learning framework for strength prediction

  • 期刊论文

  • 发表日期：

    2026年2月24日

查看全文
  • P. Kalpana,
  • Swathi Lenka,
  • Kishore Bhamidipati,
  • M. Susmitha,
  • J. Prasanya,
  • Venubabu Rachapudi,
  • Tahera,
  • Srushti V. Hosamath

摘要

Eco-friendly alternatives to expensive synthetic fibers and high-resource materials in concrete have been developing rapidly over recent years. Due to the cost-effective and renewable nature of jute fibers, they have great potential as a source of low-carbon concrete; however, their effects on the mechanical characteristics of concrete are still poorly defined. The integrated framework combining ensemble and interpretable machine learning algorithms is developed to investigate and predict the compressive strength of Jute Fiber-Reinforced Concrete (JFRC), based on a large and highly harmonized database of literature-based experimental data. Four existing Machine Learning (ML) methods, namely the Random Forest-Gradient Boosting (RF-Gb) Ensemble, AdaBoost, CatBoost, and Relevance Vector Machine (RVM), were used. It was determined that CatBoost and the RF-Gb method performed best in terms of predicting experimental results (compressive strength), and that there was excellent correlation between the predicted and actual experimental results. The SHAP-based interpretability analysis also revealed that flexural strength and tensile strength at split-tensile testing were the most significant factors affecting compressive strength, while jute-fiber dosage, workability, and replacement of fine aggregate were secondary influencing factors. These findings show the feasibility of using jute fibers as a sustainable reinforcement in the production of low-carbon concrete and that the application of interpretable ML can provide a means of supporting informed design of mixes for optimized, low-carbon, natural fiber-concrete products. Overall, the present study provides a data-driven and scalable framework that supports the development of high-performance and environmentally responsible, fiber-reinforced composites.