Prediction of rheological properties of high polymer-modified asphalt binders based on BAS-BP neural network and functional groups

High polymer-modified asphalt binder exhibits superior performance and is increasingly gaining attention. However, the literature suggests a lack of correlation analysis between their performance and chemical composition, as well as performance prediction methods. Therefore, this study aims to reveal the rheological performance of high polymer-modified asphalt binder and to develop a prediction model based on the correlation between its rheological properties and chemical composition. Polymer-modified asphalt binders with varying styrene–butadiene–styrene polymer content were prepared, and a series of dynamic shear rheometer tests and bending beam rheometer tests were conducted at high and low temperatures. Further, the chemical composition of the asphalt binders was analyzed using Fourier-transform infrared spectroscopy. Grey relational analysis was employed to examine the correlation between various performance indicators and chemical composition parameters. Finally, based on this correlation, a beetle antennae search-back propagation neural network was developed to predict its rheological performance. Rheological experiments revealed that the performance of high polymer-modified asphalt binders improved with increasing polymer content, with both high- and low- temperature performance benefiting from the combined effects of the butadiene and styrene segments in the polymer. The characteristic functional groups of butadiene and styrene showed a linear relationship with polymer content in the binder and correlated with rheological indicators, making them “phenotypic genes” of the material. The predictive model constructed in this study effectively predicted the rheological performance of high polymer-modified asphalt binder based on its chemical composition parameters. Although the data are limited, the average relative error for predictions of eight performance indicators is below 3%, and additional data will further enhance the model's accuracy and applicability. Overall, these findings provide new perspectives and methods for linking the chemical composition and rheological properties of polymer-modified asphalt binders and for performance prediction, contributing to the understanding of the asphalt material genome.