Estimation of Rutting Depth of Asphalt Mixtures Containing WMA-Rubberized Binder Using Neural Network and Nonlinear Regression Model

The use of modified bitumens has increased the number of variables affecting pavement rutting resistance. Based on this, investigating pavement rutting behavior as an effective approach of functional and environmental variables can improve operational conditions during design and maintenance and reduce heavy laboratory costs. In this research, an attempt is made to evaluate the rutting behavior of bitumen and asphalt mixtures containing the simultaneous combination of rubber powder and Sasobit using multiple stress creep recovery (MSCR) and dynamic creep tests. Then, based on the cumulative plastic strains occurring in the pavement, the rutting resistance of asphalt mixtures will be estimated using artificial neural network models and multiple regression, so that if a suitable model is determined with high accuracy and low error, the production of asphalt with high rutting potential in the laboratory stage and prevent before factory production. The results of the laboratory section indicated that despite the positive effect of GTR and WMA on the rutting resistance of bitumen and asphalt mixtures in high percentages, but taking into account the technical performance of the pavement at high and low temperatures and the economic conditions of bitumen containing 12% rubber powder and 2 % Sasobit is suggested as the optimal combination. Also, the results of the modeling section showed that despite the appropriate performance of the regression and ANN models in estimating the rutting resistance, the ANN model with a correlation coefficient of 0.939 was better than the regression model in terms of accuracy and power. Therefore, it can be suggested as a powerful and appropriate tool in reducing time and cost and preventing the production of asphalt with high rutting potential in the laboratory stage and before factory production.