Experimental Investigation and ML Based Prediction of Compressive Strength of Concrete with Substitution of Coarse Aggregates by Agro-Waste under Wax and Water Curing Regimes

Abstract

The rapid urbanisation of the Earth’s inhabited continents has led to the depletion of the planet’s natural aggregate reserves. In response, scientists and engineers have begun to seek out alternative construction materials that are both sustainable and capable of maintaining the structural integrity of constructed buildings. One such material is coconut-shell-based Agro-Waste (AW). The present investigation studied the impact of using AW to replace the coarse aggregates in M25-grade concrete at 10% and 20% replacement rates by volume, using both conventional water and wax curing methods. “A total of 54 cube specimens were cast, representing six mix-curing combinations tested at 7, 14, and 28 days with three replicate specimens per condition.” Results indicated that replacing 10% of the coarse aggregates with AW and curing the resultant concrete in wax yielded a higher compressive strength (29.58 MPa at 28 days) than standard concrete that employed water curing (24.76 MPa) - a difference of 19.5%. However, using 20% replacement of coarse aggregates with AW resulted in a reduction in the strength of the resultant concretes relative to the reference standard. An Artificial Neural Network (ANN) was developed to predict the compressive strength of concretes made with various percentages of AW and cured with different methods. Using four different features, the ANN achieved an R² value of 0.9847, with a Root-Mean-Squared Error (RMSE) of 0.287 MPa and a Mean-Absolute-Error (MAE) of 0.219 MPa - outperforming Random Forest, XGBoost, and Support Vector Regression algorithms. The features that contributed most significantly to the ANN’s prediction power were the age of the concretes and the percentage of AW used. This study indicates that using AW in place of coarse aggregates and curing wax offers environmental benefits to both land and water resources in tropical countries that experience both landfill and water scarcity problems. These findings have the potential to inform the construction of low-rise buildings in resource-limited environments. The strength results were analysed using the mean value, standard deviation, coefficient of variation and analysis of variance to determine the significance of the agro-waste content and curing regime on the strength of the products. Within the limited experimental data of coconut shell replacement in the range of 0 to 20 percentage and curing age of 7 to 28 days, the ANN exhibited a strong ability to interpolate.