Litcius/Paper detail

Predicting the Compressive Strength of Waste Powder Concrete Using Response Surface Methodology and Neural Network Algorithm

Hany A. Dahish, Mohammed K. Alkharisi, Mohamed A. Abouelnour, Islam N. Fathy, Marwa A. Sadawy, Alaa A. Mahmoud

2025Buildings7 citationsDOIOpen Access PDF

Abstract

The rapid development in building construction has stimulated the replacement of cement in concrete with construction waste materials such as marble waste powder (MWP) and granite waste powder (GWP) to reduce the negative impact of cement production and to save natural resources. Therefore, the inclusion of these materials in concrete contributes to environmental sustainability by reducing cement consumption and promoting the reuse of industrial waste. The present study employs Response Surface Methodology (RSM) and, for the first time in a comparable context, the Neural Network Algorithm (NNA) as an advanced optimization and predictive tool to evaluate the synergistic effect of using GWP and MWP as partial cement replacements in concrete exposed to elevated temperatures. The study involved four independent variables: replacement level of GWP up to 9%, replacement level of MWP up to 9%, the degree of temperature (T) up to 800 °C, and the exposure duration (D) up to 2 h, while the dependent variable (output) was the compressive strength (CS). The ANOVA results revealed that the quadratic model outperformed the linear model in predicting the CS of concrete. The Quadratic model, derived from RSM, demonstrated superior performance in predicting CS values. However, the NNA model also showed high predictive accuracy (R2 = 0.949; RMSE = 1.5297 MPa), effectively capturing the complex and nonlinear relationships among temperature, duration, and the cement replacement levels with GWP and MWP. The optimization results revealed that the maximum compressive strength of 39.4 MPa can be achieved at 8.92% GWP, 1.89% MWP, T of 247 °C, and D of 0.64 h with a desirability of 1. The proposed models provided valuable insights into the synergistic effects of granite and marble waste powders, supporting the design of sustainable, high-performance concrete with reduced environmental footprint and improved resource efficiency.

Topics & Concepts

Response surface methodologyCompressive strengthCementArtificial neural networkReuseMaterials scienceEnvironmental scienceDesign of experimentsNonlinear systemStructural engineeringQuadratic equationProperties of concreteComputer scienceMunicipal solid wasteNonlinear programmingSuperplasticizerSensitivity (control systems)Construction industryAlgorithmWaste managementProcess engineeringIndustrial wasteUltimate tensile strengthEnergy consumptionCentral composite designQuadratic modelProduction (economics)Concrete and Cement Materials ResearchFire effects on concrete materialsRecycled Aggregate Concrete Performance
Predicting the Compressive Strength of Waste Powder Concrete Using Response Surface Methodology and Neural Network Algorithm | Litcius