Identification of chemical markers for blending phenomena in RAP mastics using FTIR spectroscopy and multivariate discriminant analysis

This study investigates the complexity associated with the blending phenomenon within recycled asphalt pavement (RAP) binder by employing Fourier Transform Infrared (FTIR) spectroscopy in conjunction with multivariate discriminant analysis, specifically utilizing Partial Least Squares Regression (PLSR) and Linear Discriminant Analysis (LDA) algorithms, collectively termed PLSR-LDA approach. The dataset comprises mastics with varying RAP contents, ranging from 0 % to 100 %, and includes two warm mix additives, Aspahmin and Sasobit. The multivariate discriminant analysis using PLSR-LDA effectively differentiated bituminous mastics based on RAP concentrations and warm mix additives, identifying significant wavenumbers with high normalized variance importance in projection (VIP) scores. Based on these normalized VIP scores, RAP concentrations were accurately identified by the absorption peaks at 1200–1145 cm −1 (aliphatic C–H bending vibrations) and 836–785 cm −1 (C–H bending vibrations in long-chain hydrocarbons) achieving an overall accuracy of nearly 80 %. In contrast, warm mix additives were effectively classified using the peaks at 735–713 cm −1 (Sasobit marker), 1135–900 cm −1 (Asphamin marker), and 2945–2875 cm −1 (C–H stretching in aliphatic hydrocarbons) with an accuracy exceeding 95 %. To further distinguish between different bitumen classes, the CatBoost classifier was employed, evaluating the absorption ranges based on feature importance values. The highlighted absorption peaks showed that RAP inclusion in bituminous mastics introduces aging-induced chemical components into the blended binder, such as aliphatic C–H bending vibrations and oxygen-containing compounds. This observation suggests that the blending between RAP binder and virgin bitumen can be chemically tracked using FTIR spectroscopy and multivariate discriminant analysis, proposing this approach for gauging the blending degree in future works. • PLSR-LDA approach effectively classified bituminous mastics based on RAP content and the inclusion of warm mix additives. • Wavenumbers at 1454, 1305, 1597, and 1699 cm-1 were identified as critical to analyzing the RAP blending phenomenon. • WMA additive inclusion was successfully identified through the Sasobit and Asphamin markers. • The research revealed that increased RAP content introduces higher aging markers into the blended binder.