Asphalt Quality Evaluation and Intelligent Classification via ATR-FTIR Functional Group Analysis

Conventional asphalt testing methods inadequately and slowly represent the inherent chemical composition variations among materials. This study utilizes attenuated total reflectance Fourier transform infrared spectroscopy (ATR-FTIR) to identify subtle changes in functional groups within asphalt binders. Additionally, it integrates machine learning technology to develop a rapid and intelligent framework for quality evaluation and brand recognition based on chemical fingerprints. This paper examines four distinct brands of No. 70 needle penetration grade (PEN 70) base asphalt binders, with spectral information acquired through ATR-FTIR analysis. This work innovatively links the characteristic differences of functional groups, as captured by FTIR, to asphalt brand affiliation and quality characteristics, enabling intelligent and high-precision classification via machine learning. Principal component analysis (PCA) indicated that various asphalt brands exhibit notable differences in spectral characteristics within the fingerprint region (1,300–400 cm−1) and in the relative content of aromatic and aliphatic functional groups, which is essential for brand differentiation. Characteristic functional group peaks at 1,376, 813, 865, and 722 cm−1 were identified as essential for brand differentiation. The differences in functional groups serve to differentiate brands and correlate with performance indicators, including ductility. Random forest (RF) is the most effective machine learning algorithm for learning patterns of functional group changes. The optimized RF model attained an overall classification accuracy of 98.4% on an independent test set, as determined by fingerprint region spectral data. This framework offers a dependable technical method for the rapid quality assurance of asphalt in location.