Scholarlink

<p>In structural engineering, concrete compressive strength (CS) is among its most essential performance characteristics. Although many machine learning models have been developed for predicting this parameter, many suffer from limited transparency. This study developed an accurate and explainable ML model based on the Levenberg–Marquardt algorithm for estimating concrete CS. A total of 1030 laboratory measured concrete CS data points was used for model development. Model performance was evaluated using the coefficient of determination (R<sup>2</sup>), root mean square error (RMSE), average absolute percentage relative error (AAPRE), and average percentage relative error (APRE). For the test dataset, the model yielded 0.949 for R<sup>2</sup>, 3.77&#xa0;MPa for RMSE, 10.32% for AAPRE and − 1.92% for APRE. Sensitivity analysis identified the binder proportion as the most contributing variable with a factor of + 0.52. This is trailed by the amount of superplasticizer (+ 0.39) and the age of curing the sample (+ 0.35). Moreover, the proposed model is presented in an explicit mathematical form for straightforward integration into relevant software; an attribute rarely addressed in existing ML based studies. The physical trend analysis confirmed consistency with established concrete strength behaviour. Finally, the development of a user friendly graphical user interface for the framework facilitates easy deployment of the model for rapid estimation of concrete CS.</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

Explainable Levenberg Marquardt trained neural network paradigm for forecasting concrete compressive strength

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年7月1日

查看全文
  • Okorie Ekwe Agwu,
  • Rubayyi Turki Alqahtani,
  • Etieno Bassey Udofia,
  • Muhammad Aslam Md Yusof

摘要

In structural engineering, concrete compressive strength (CS) is among its most essential performance characteristics. Although many machine learning models have been developed for predicting this parameter, many suffer from limited transparency. This study developed an accurate and explainable ML model based on the Levenberg–Marquardt algorithm for estimating concrete CS. A total of 1030 laboratory measured concrete CS data points was used for model development. Model performance was evaluated using the coefficient of determination (R2), root mean square error (RMSE), average absolute percentage relative error (AAPRE), and average percentage relative error (APRE). For the test dataset, the model yielded 0.949 for R2, 3.77 MPa for RMSE, 10.32% for AAPRE and − 1.92% for APRE. Sensitivity analysis identified the binder proportion as the most contributing variable with a factor of + 0.52. This is trailed by the amount of superplasticizer (+ 0.39) and the age of curing the sample (+ 0.35). Moreover, the proposed model is presented in an explicit mathematical form for straightforward integration into relevant software; an attribute rarely addressed in existing ML based studies. The physical trend analysis confirmed consistency with established concrete strength behaviour. Finally, the development of a user friendly graphical user interface for the framework facilitates easy deployment of the model for rapid estimation of concrete CS.