Correlation between the Molecular Structure and Viscosity Index of CTL Base Oils Based on Ridge Regression

Abstract

In China, coal-to-liquid (CTL) lube base oils with ultrahigh viscosity index (VI) are very popular. Since it consists of chain alkanes only and can be precisely characterized by molecular structures alone, quantitative 13C nuclear magnetic resonance (NMR) data are used to generate the average structural parameters (ASPs) of CTL base oil. In this work, the ASPs and bulk properties of CTL base oils were tested and compared with those of mineral base oils. Based on the test results, the correlation between the unique property of CTL base oil VI and ASPs was analyzed. To eliminate the effect of significant multicollinearity among the input variables, statistical methods such as ordinary least-squares (OLS), stepwise regression, and ridge regression methods were used to build the VI prediction model. The main findings are as follows: according to the 13C NMR spectrum, CTL base oils had a significantly higher content of isomeric chain alkanes (including several branching structures) than mineral base oil, while the content of cycloalkanes was zero; among several branched structures, the one with the largest difference in content is structure S67, which has the highest percentage in the iso-paraffin structures, all above 25.5% in CTL base oils and below 21.39% in mineral oils; according to the distillation curve of the simulated distillation (SimDist) analysis, CTL base oils with similar carbon number distribution showed lower boiling points, narrower distillation ranges, and higher distillation efficiencies than mineral base oil; correlation analysis showed that the average chain length (ACL), normal paraffins (NPs), and structure S67 caused the CTL base oil to exhibit a higher VI; and from 13C NMR data, the ridge regression model was used to obtain regression coefficients consistent with reality, and the expected VI could be well predicted with a correlation coefficient of 0.935.