Concreto estrutural de cimento Portland com incorporação de agregado miúdo reciclado: avaliação e modelagem computacional

Abstract

The growing concern over the environmental impacts generated by the construction industry has encouraged research focused on the reuse of waste materials in the production of structural concrete. River sand extraction and the improper disposal of waste from concrete products constitute serious environmental problems, contributing to the degradation of water bodies, biodiversity loss, and the increase in urban environmental liabilities. In this context, the present work comprises two complementary studies addressing the use of recycled fine aggregate derived from industrial waste generated by precast concrete products. The first article presents the results of an extensive experimental program involving the production of concrete mixtures with different replacement levels of natural sand by recycled fine aggregate. The mechanical properties evaluated were axial compressive strength and splitting tensile strength. Statistical analysis was performed using analysis of variance (ANOVA), the Scott–Knott test, and regression model fitting. For replacement levels of 10%, 20%, 30%, 40%, and 50%, the average increases in compressive strength were 1.69%, 6.69%, 6.30%, 8.44%, and 9.05%, respectively, whereas the average variations in tensile strength were 0.84%, −0.53%, 0.50%, 0.86%, and 1.78%. These results indicate that, on average across the evaluated mixtures, the incorporation of recycled aggregate did not significantly compromise the mechanical performance of the concrete. The regression models yielded coefficients of determination (R²) greater than 0.85 for most of the fitted models. The second article proposes an artificial intelligence-based predictive approach to estimate the proportions of the concrete constituent materials, considering the recycled aggregate replacement percentage and the required mechanical strengths as input variables. Several Artificial Neural Network (ANN) models and Fuzzy Inference Systems (FISs) were developed and compared. The ANNs performed better in terms of error metrics, including root mean square error (RMSE) and mean absolute error, whereas the FISs achieved better coefficients of determination, with R² values greater than 0.9. The integration of both studies strengthens the potential use of recycled aggregates in the production of structural concrete while demonstrating the effectiveness of computational intelligence tools for predicting concrete mix proportions.