Rapid analysis of complex components in residual oil based on near-infrared spectroscopy combined with chemometrics

Residual oil, a significant byproduct of petroleum refining, requires analysis of its complex components to optimize refining processes and meet ecological regulatory requirements. However, traditional methods for analyzing residual oil components are time-consuming and incur high environmental costs. In contrast, NIR spectroscopy offers a rapid, non-destructive, and environmentally friendly alternative for efficiently detecting complex components such as sulfur content (SC), coking value (CV), and SARA (saturates, aromatics, resins, and asphaltenes) fractions. Due to the distinct response patterns exhibited by residual oil samples in NIR spectra resulting from varying refining processes, this study employs Fuzzy C-Means (FCM) clustering to categorize samples. Through FCM, a subset of core samples with highly consistent chemical properties was identified for subsequent modeling. By integrating spectral preprocessing, Competitive Adaptive Reweighted Sampling (CARS), and Partial Least Squares (PLS) regression, an efficient NIR quantitative analysis model was developed for the determination of residual oil components. The model exhibited excellent performance, with a coefficient of determination (R² > 0.89) and residual predictive deviation (RPD > 3.0) for the validation set. Root mean square errors of prediction (RMSEP) and corresponding concentration ranges (% w/w) were as follows: sulfur content 0.14 (0.37–4.47), coking value 0.73 (6.67–18.14), saturates 1.35 (18.29–41.39), aromatics 3.27 (26.62–62.90), resins 4.32 (1.36–44.79), and asphaltenes 0.34 (1.92–8.37). Paired t-tests (p > 0.05) confirmed no significant differences between predicted and actual values.