Behavior Assessment, Regression Analysis and Support Vector Machine (SVM) Modeling of Waste Tire Rubberized Concrete

Abstract An experimental study was undertaken to assess the properties of rubberized concrete made by recycled tire rubber. Mechanical tests were carried out and the ultrasonic technique was also used to measure the acoustic properties of the rubberized concrete, such as wave velocity, and the relationships between the mechanical properties and pulse velocity for different mixtures were investigated. Then, based on the experimental data, a thorough strength modeling was performed using regression analysis and support-vector machine (SVM) technique. To develop the predictive models for strength behavior of the recycled rubber concrete, a comprehensive regression analyses including ANOVA, t-test and F-test were conducted, and the significance of the influencing variables was determined. Then, several regression models by incorporating different combinations of the variables were developed and compared. Finally, the SVM method as a machine-learning technique was utilized to develop several models by using different kernel functions, optimization algorithms, and hyperparameters optimization to predict the compressive strength of the recycled rubber concrete. The results obtained were compared with those of regression models and it was found that SVM outperforms all the regression models assessed in this study.