Scholarlink

<p>This study investigates the effect of plastic optical fiber integration on the compressive strength of M20 concrete. SVM-RBF, SVM-Linear, and XGBoost based machine learning prediction models were also trained and compared with conventional linear regression model. The compressive strength analysis confirms that POF inclusion reduces strength due to weak interfaces and void formation, particularly at smaller fiber spacings. However, increasing fiber spacing to 20 mm minimizes strength loss, demonstrating a more viable configuration for practical applications. The performance metrics, regression error characteristic (REC) curves, taylor diagram, and area over curve (AOC) results highlight XGBoost as the most accurate predictive model, outperforming SVM-RBF, SVM-Linear, and linear regression models. The R<sup>2</sup> values in training and testing for the XGBoost model are 0.999 and 0.997 respectively. The RMSE values in training and testing for the XGBoost model are 0.151 and 0.259 respectively. The monotonicity analysis reveals that fiber spacing and curing days positively affect compressive strength, while other mix variables remain relatively unchanged within the tested range.</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

Machine learning-based prediction of compressive strength of plastic optical fiber embedded transparent concrete

  • 期刊论文

  • 发表日期：

    2025年4月18日

查看全文
  • Manish Pratap Singh,
  • Anish Kumar,
  • Sanjeev Sinha

摘要

This study investigates the effect of plastic optical fiber integration on the compressive strength of M20 concrete. SVM-RBF, SVM-Linear, and XGBoost based machine learning prediction models were also trained and compared with conventional linear regression model. The compressive strength analysis confirms that POF inclusion reduces strength due to weak interfaces and void formation, particularly at smaller fiber spacings. However, increasing fiber spacing to 20 mm minimizes strength loss, demonstrating a more viable configuration for practical applications. The performance metrics, regression error characteristic (REC) curves, taylor diagram, and area over curve (AOC) results highlight XGBoost as the most accurate predictive model, outperforming SVM-RBF, SVM-Linear, and linear regression models. The R2 values in training and testing for the XGBoost model are 0.999 and 0.997 respectively. The RMSE values in training and testing for the XGBoost model are 0.151 and 0.259 respectively. The monotonicity analysis reveals that fiber spacing and curing days positively affect compressive strength, while other mix variables remain relatively unchanged within the tested range.