Experimental evaluation and predictive modeling of sustainable concrete incorporating agricultural and industrial waste admixtures

Abstract

The sustainable use of agricultural and industrial waste materials in concrete requires quantified experimental evidence and interpretable prediction tools for engineering decision-making. This study evaluated concrete containing sugar, cow bone ash, groundnut shell ash and limestone powder and developed interpretable strength models. A 1:2:4 mix used 317 cementitious material, 739 fine aggregate and 1380 coarse aggregate at 0.50 water-to-binder ratio, with 0 to 30% admixture by cement mass. Tests measured slump, initial/final setting time, 7-, 14- and 28-day strength, water absorption and drying shrinkage. The control reached 27.1 MPa at 28 days. Cow bone ash at 8% reached 31.2 MPa, a 15.1% increase while limestone powder at 10% reached 33.3 MPa, a 22.9% increase. These optimum mixtures reduced water absorption from 4.5% to 3.4% and 3.2% and drying shrinkage from 0.050% to 0.043% and 0.038%, respectively. Groundnut shell ash at 10% reduced 28-day strength to 19.5 MPa while sugar at 2% reduced it to 24.5 MPa. Modelling used 144 complete observations screened from 160 records. The artificial neural network achieved test = 0.9877, RMSE = 0.8966 MPa and MAE = 0.7132 MPa; the adaptive neuro-fuzzy inference system achieved = 0.9190, RMSE = 2.3012 MPa and MAE = 1.9406 MPa. 5-fold cross-validation gave = 0.966, RMSE = 1.80 MPa and MAE = 1.27 MPa. Residual diagnostics, Shapley feature influence, partial dependence analysis and an uncertainty band of ± 2 MPa supported interpretation. A web-based model implementation tool was also deployed to support practical preliminary mix assessment. Controlled cow bone ash and limestone powder contents improved strength and durability, while the deployed prediction workflow supports preliminary engineering mix assessment within the tested domain.