Investigation of mechanical behavior of mortar using slag as partial replacement of sand based on experimental and machine learning approaches

Non-ground granulated blast furnace slag (NGGBFS), a by-product of the iron industry, management and disposal have been identified as critical issues for industries in several nations. Ground granulated blast furnace slag (GGBFS) a modified form of NGGBFS, can effectively replace some of the fine aggregates in mortar. In the first section of this study, the experimental work on mortar in which natural sand was partially substituted with GGBFS is presented. Eight different ratios of slag (0%, 10%, 20%, 30%, 40%, 50%, 60%, and 70%) were substituted for sand in the fabrication of mortar specimens. In addition, cement-to-sand ratios of 1:2.25, 1:2.75, and 1:3.50 were employed in three different types of mortar specimens. A total of 240 cube and briquette mortar specimens were cast, and their compressive and tensile strengths were assessed at curing ages of 7, 14, 28, 60, and 90 days. The test findings demonstrate that the mortar strength (both compressive and tensile strengths) improves proportionally with the GGBFS content up to an optimal level (30%), beyond which the strength deteriorates with further GGBFS addition. The compressive and tensile strengths of mortar are improved by 22% and 18%, respectively, at the optimum value compared to the reference mortar. As there is no proven prediction model, an artificial neural network (ANN) has been deployed in the second section to forecast the mechanical characteristics of the tested specimens. The established ANN model is capable of predicting test results rather accurately.