Experimental Evaluation and Modeling of Polymer Nanocomposite Modified Asphalt Binder Using ANN and ANFIS

In this study, six different blends including a base binder, polymer modified binder [5% acrylonitrile styrene acrylate (ASA) by the weight of binder], ASA–nano calcium, and ASA–nano copper at concentrations of 3% and 5% by weight were characterized using a dynamic shear rheometer (DSR). The range of temperatures was $46°\mathrm{C}–82°\mathrm{C}$, while the frequencies were from 1 to $100\mathrm{rad}$/s. The prediction of complex modulus ($G^{*}$) from physical and rheological properties of binders and the mechanical test conditions were performed using an artificial neural network (ANN) and adaptive neuro-fuzzy inference system (ANFIS). The coefficient of determination ($R^2$) and root-mean-squared error (RMSE) were used as the performance indicator metrics in the evaluation of the performance of analytical models. The results of this study show improvement in the rheological behavior of the modified asphalt binder. Further, ANN and ANFIS models for predicting the outcomes of the DSR test results have been shown to provide reliable models both with training and testing data sets. The $R^2$ values of 0.996 and 0.920 and RMSE values of 0.008295 and 0.006755 were obtained with the testing data sets for the ANN and ANFIS prediction models, respectively. Model results showed that both ANN and ANFIS models were able to predict $G^{*}$ with high accuracy, with ANN being the more efficient analytical model in terms of performance.