Artificial neural networks (ANN), MARS, and adaptive network-based fuzzy inference system (ANFIS) to predict the stress at the failure of concrete with waste steel slag coarse aggregate replacement

Concrete is a very flexible composite material that is extensively employed in the building industry. Steel slag is a waste material produced during steelmaking. It is formed during the separation of molten steel from impurities in steelmaking furnaces. Slag starts as a molten liquid melt and cools to a solid state. It is a solution of silicates and oxides that is rather complicated. Steel slag recovery is environmentally friendly since it conserves natural resources and frees up landfill space. Steel slag has been extensively utilized in concrete as a partial substitute for normal and crushed coarse aggregate to improve the mechanical qualities of normal-strength concrete, such as compressive strength. The researchers and suppliers investigated that using steel slag instead of normal coarse aggregate could save the environment and natural resources. Three hundred thirty-eight (338) data sets were gathered and evaluated in total. During the modeling procedure, the most significant factors affecting the compressive strength of concrete with steel slag replacement were considered, including the curing time of 1-180 days, the cement content of 237.35-550 kg/m(3), the water-to-cement ratio of 0.3-0.872, the fine aggregate content of 175.5-1285 kg/m(3), the steel slag content of 0-1196 kg/m(3), and the coarse aggregate content of 0-1253.75 kg/m(3). A credible mathematical model is needed to investigate the influence of steel slag as a partial replacement on concrete compressive strength. Mathematical models will help engineers and concrete industries mix a proper concrete mix design, including steel slag, to achieve a desired compressive strength without doing any experimental work. As a result, an artificial neural network (ANN), an adaptive network-based fuzzy inference system (ANFIS), a multivariate adaptive regression splines (MARS), and an M5P-tree model were presented in this research to predict the compressive strength of concrete with steel slag aggregate replacement. According to previous research findings, all percentages of steel slag improve compressive strength. According to statistical studies, the adaptive network-based fuzzy inference system model outperformed the other models in forecasting steel slag replacement compressive strength for normal strength concrete (ANN, MARS, and M5P-tree). It has a higher coefficient of determination of 0.99, a smaller mean absolute error of 0.74 MPa, a smaller root mean square error of 1.12 MPa, a smaller scatter index of 0.029, and a smaller objective of 0.93 MPa.