Implementation of Regression Analysis and Artificial Neural Network in the Prediction of Rubberized Concrete Mechanical Properties

Recent developments in the field of construction materials established the foundation for the usage of rubberized concrete as a structural material due to its high seismic damping performance. Another objective of the rubberized concrete is to reduce the footprint of wasted rubber by utilizing it as partial aggregate replacement in concrete mixture. Various studies have been conducted to characterize the engineering properties of rubberized concrete. However, none of which presented a generalized model which can be used worldwide to obtain various concrete grades with different rubber replacement percentages. In this study, a comprehensive dataset is collected from over 40 research papers of the published work in the literature. Mathematical predictive models of the engineering properties of rubberized concrete are constructed on the basis of regression analysis and artificial intelligence. Results indicated that regression analysis moderately estimated the engineering properties of rubberized concrete where the coefficient of determination ranged between 0.55 and 0.8. On the other hand, the constructed model through artificial neural network has higher prediction accuracy with a coefficient of determination ranging between 0.82 and 0.96. In addition, this research presented a formula that correlate the compressive strength of rubberized concrete to its static elasticity modulus. Keywords: artificial neural network, compressive strength, splitting tensile strength, elasticity modulus, flexural strength, regression analysis, rubberized concrete.