ABSTRACT The aging phenomenon in bituminous mastics leads to distinct chemical changes that affect their mechanical and rheological properties. Experimental testing like Fourier Transform Infrared (FTIR) spectroscopy and Dynamic Shear Rheometer (DSR) often require advanced post-processing to reveal underlying patterns. This research employs advanced pattern recognition algorithms and chemometric analyses to investigate aging-induced chemical changes in bituminous mastics. The study seeks to establish a rational relationship between their chemical and rheological properties. FTIR data were analyzed using Partial Least Squares Regression-Linear Discriminant Analysis (PLSR-LDA) to derive normalised Variable Importance in Projection (VIP) scores. Post-processing of PLSR-LDA results, including Pearson correlation testing and classification through data mining algorithms, identified carbonyl groups, alcohols, and carboxylic acids as key indicators of aging. Significant mineral vibrations in the Si-O band and carbonate-related vibrations were also important for differentiating various mineral compositions in the mastics. The study demonstrated that combining PLSR with CatBoost regression analysis effectively predicts rheological behaviour based on chemical composition. These findings underscore the non-linear relationship between their chemical and rheological properties and provide a framework for predicting material performance based on chemical characteristics.
Read full abstract