Characterisation of bitumen grade and modification using a multilevel multivariate discriminant analysis approach and FTIR spectroscopy

The complex nature of bitumen poses significant challenges in its characterisation. This study employed FTIR spectroscopy combined with Partial Least Squares Regression and Linear Discriminant Analysis algorithms (PLSR-LDA) to classify bitumen samples based on chemical composition. The dataset included samples from multiple manufacturers, with varying ageing levels and the incorporation of SBS and wax. The multilevel PLSR-LDA approach effectively differentiated bitumen grades, SBS modification levels and RAP binders by identifying key chemical markers. Unmodified bitumen samples were classified based on C-H bending and stretching in aliphatic compounds, as well as CH₂ rocking vibrations. In SBS-modified samples, distinctions were driven by variations in aromatic and aliphatic components linked to SBS and wax. For RAP binders, classification was governed by CH₂ rocking vibrations, oxidative ageing markers (sulphoxide, carbonyl) and changes in aliphatic CH₂/CH₃ structures.