Prediction of the Compressive Strength of Rubberized Concrete Based on Machine Learning Algorithm

Abstract

In recent years, car tires are considered one of the most crucial environmental pollution problems in many countries. Therefore, reusing waste rubber crumbs from recycled tires as aggregates for concrete has attracted increasing attention. Rubberized concrete could also be an economical and environmentally-friendly construction material. Besides, enhancing the ductility, toughness, thermal insulation, and impact resistance are also advantageous while using rubberized concrete. On the contrary, rubberized concrete’s mechanical properties are highly dependent on the replacement amount of rubber. Thus, the estimation of the compressive strength of rubberized concrete is crucial for engineering applications. In this study, 162 experimental results collected from the literature are used to construct a database and attempt to predict the compressive strength of rubberized concrete. An artificial neural network (ANN) is developed, using 7 input variables, namely binder, superplasticizer, water, fine aggregate, coarse aggregate, crumb rubber, and chipped rubber. The model performance is evaluated using three performance indicators, such as root mean square error (RMSE), mean absolute error (MAE), the Pearson correlation coefficient (R). The results show that the proposed ANN algorithm exhibits excellent prediction performance and accurately estimated the compressive strength of rubberized concrete. The results in the present research are useful and could provide a reference for engineers in predicting the compressive strength of rubberized concrete.