Abstract
This study explores the application of multivariate data analysis in the viscosity prediction of crude oils using NMR relaxation data. The 1H transverse relaxation times ( T 2) of 68 Brazilian crude oil samples, ranging from light to extra-heavy (2 to 30,000 cP), were measured at 2 MHz. Partial least squares regression (PLSR) models were developed to predict the oil viscosity in log viscosity units from the T 2 relaxation spectra and directly from the raw relaxation curves. In both cases, the PLSR with only three latent variables produced good calibration models, with a standard error of prediction of 0.161 and 0.135 log cP for the T 2 relaxation spectra and raw relaxation curves, respectively. The PLSR models were validated by full cross and external set schemes revealing quite equivalent performances.
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have