Scholarlink

<p>This study compares the effectiveness of Multiple Linear Regression (MLR) and Multilayer Perceptron (MLP) models in predicting the compressive strength of ceramic waste-based concrete, addressing the need for accurate strength prediction in sustainable concrete development. A dataset of 177 concrete mix designs, encompassing various concrete types and curing ages, was utilized. The MLR model achieved an R<sup>2</sup> of 83.7% and RMSE of 7.17&#xa0;MPa, providing interpretable coefficients for each predictor. The MLP model, featuring two hidden layers with dropout regularization, slightly outperformed MLR with an R<sup>2</sup> of 85% and RMSE of 5.71&#xa0;MPa, potentially capturing more complex relationships between variables. Both models demonstrated good generalization ability through cross-validation and early stopping techniques. While MLP showed marginally superior predictive accuracy, MLR offered greater interpretability of variable contributions. The study highlights the positive impact of ceramic waste powder on compressive strength, supporting its use as a sustainable concrete additive.</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

Predicting Compressive Strength of Ceramic Waste-Based Concrete: A Comparative Study of MLP and Regression Models

  • 期刊论文

  • 发表日期：

    2025年9月5日

查看全文
  • Amit Mandal,
  • Sarvesh P. S. Rajput

摘要

This study compares the effectiveness of Multiple Linear Regression (MLR) and Multilayer Perceptron (MLP) models in predicting the compressive strength of ceramic waste-based concrete, addressing the need for accurate strength prediction in sustainable concrete development. A dataset of 177 concrete mix designs, encompassing various concrete types and curing ages, was utilized. The MLR model achieved an R2 of 83.7% and RMSE of 7.17 MPa, providing interpretable coefficients for each predictor. The MLP model, featuring two hidden layers with dropout regularization, slightly outperformed MLR with an R2 of 85% and RMSE of 5.71 MPa, potentially capturing more complex relationships between variables. Both models demonstrated good generalization ability through cross-validation and early stopping techniques. While MLP showed marginally superior predictive accuracy, MLR offered greater interpretability of variable contributions. The study highlights the positive impact of ceramic waste powder on compressive strength, supporting its use as a sustainable concrete additive.