Predicting rutting performance of carbon nano tube (CNT) asphalt binders using regression models and neural networks

The complex behavior of asphalt binders makes it difficult to accurately predict their complex modulus (G*) and rutting performance (G*/Sin (δ)). The aim of this study was to investigate the effects of loading frequency and temperature on rutting susceptibility of CNT asphalt binders. To predict the rutting performance of a CNT-modified binder, two techniques, i.e. regression models and artificial neural networks (ANN), were used. The proposed artificial neural network received CNT content, test temperature and loading frequency as the input and provided the complex modulus as the output. Totally, 480 combinations were evaluated. To test the effects of CNT content and mechanical properties on the rutting performance of the modified binders, the Response Surface Method was used. The results showed that the ANN technique performed better in predicting the rutting performance than regression models. R2 values were 0.997, 0.819, and 0.420 in ANN, multiple regression, and linear regression, respectively. ANOVA tests showed that temperature, loading frequency and CNT percentage had a significant effect on complex modulus and rutting performance of the binder. In fact, CNTs enhanced the rutting performance and rheological behavior of the asphalt binder.