For the first time,
the development of prediction models of the penetration grade and the softening point
of vacuum residues (VRs) and pavement asphalts, from the structural
data obtained with proton nuclear magnetic resonance (1H NMR) and relaxometry data obtained via low-field nuclear magnetic
resonance (LF NMR), is reported. The correlation between the structural
data (1H NMR, percentage of different proton kinds), the
relaxometry data (T
2, spin–spin
relaxation time), and the properties, was measured with principal
component regression (PCR). The best models were those obtained with
PCR, which were validated via k-fold cross-validation,
with k = 10. In particular for the VR, the best model
for the penetration grade was obtained from LF NMR, with a training R
2 of 0.99 and a validation R
2 of 0.96; the best softening point was obtained from
the combination of 1H NMR and LF NMR, with R
2 values of 0.99 and 0.87, respectively. For the asphalts,
the best model for the penetration grade was also obtained from the
combination of 1H NMR and LF NMR, with R
2 values of 0.99 and 0.94, respectively. Note that these
prediction methods require less sample quantity, time, and personal
effort than the ASTM standards.