Investigating the differentiation of bitumen samples based on FTIR spectra using multivariate analysis (PCA-LDA) in the RILEM 295-FBB TG1 round robin test
Jan Unterbuchschachner (Technische Universität Wien)
Sandra Weigel (Federal Institute for Materials Research and Testing Berlin)
Mohsen Motevalizadeh (University of Kassel)
Sadaf Khalighi (TU Delft - Civil Engineering & Geosciences)
Johannes Mirwald (Technische Universität Wien)
Aikaterini Varveri (TU Delft - Civil Engineering & Geosciences)
Bernhard Hofko (Technische Universität Wien)
More Info
expand_more
Other than for strictly personal use, it is not permitted to download, forward or distribute the text or part of it, without the consent of the author(s) and/or copyright holder(s), unless the work is under an open content license such as Creative Commons.
Abstract
Attenuated Total Reflection Fourier Transform Infrared (ATR-FTIR) spectroscopy is a widely applicated method in bituminous binder analysis. However, differences in equipment, measurement routines, sample preparation, device and spectral evaluation make comparability of results between laboratories difficult. To demonstrate these difficulties, the task group 1 of the RILEM TC 295-FBB “Fingerprinting bituminous binders using physico-chemical analysis” has conducted an international round robin test. A total of 6,786 spectra were obtained from 21 laboratories. The laboratories used three different unmodified 70/100 penetration grade binders at three aging states and prepared them for measurement with the ATR-FTIR spectroscope using six different methods. Building on this dataset the present study applies multivariate analysis (MVA) to further investigate spectral variability and assess its potential for robust classification. Using min–max normalization and z-score filtering, the data was prepared for evaluation with Coefficient of Variation, Intraclass Correlation Coefficient and Principal Component-Linear Discriminant Analysis. Homogenized preparation methods achieved the highest reproducibility and discrimination accuracy, while solvent based and non-homogenized methods introduced greater variability. Differences between ATR-FTIR devices and manufacturers were also evident, mainly through spectral intensity shifts. Although the differences between the binders and aging states are small, an accuracy of over 0.95 demonstrates that MVA methods can clearly distinguish between them. A feature importance analysis using a random forest highlighted the carbonyl, sulfoxide, and aliphatic aromatic bands as the main factors contributing to this discriminability. This study demonstrated that ATR-FTIR spectroscopy in combination with MVA can support the development of standardized protocols for bitumen analysis.
help_outline