Main content area

Kriging Modeling to Predict Viscosity Index of Base Oils

Da Costa, J. J., Chainet, F., Celse, B., Lacoue-Nègre, M., Ruckebusch, C., Caillol, N., Espinat, D.
Energy & fuels 2018 v.32 no.2 pp. 2588-2597
confidence interval, distillation, feedstocks, kriging, liquids, models, oils, petroleum, prediction, refiners, refractive index, regression analysis, viscosity
Predicting petroleum products’ properties, such as the viscosity index (VI) of base oils, is an important challenge for refiners because the production always requires more time-consuming and costly experiments. Base oils can have very different characteristics depending on which production process they have undergone. In this work, kriging is proposed to predict the VI of base oils obtained from hydrotreatment or/and hydrocracking processes using global properties (density, refractive index, distillation curve, etc.) of feedstock and/or total liquid effluent, with conversion rate of the 370+ cut. Kriging is an interpolation method that predicts the value at a given point by computing a weighted average of the observations in the neighborhood of this point. As kriging is closely related to regression analysis, the results obtained were compared with multilinear regression (MLR). Results show that kriging and MLR have very close performances for hydrotreatment data (base oil with viscosity indices ranging from 9 to 113) for which 63% of the validation samples are predicted within the confidence interval of the standard measure. For hydrocracking data (base oils with viscosity indices ranging from 85 to 126), kriging provides better results as 62% of the validation samples are predicted in the confidence interval of the standard measure against 46% for MLR. In light of these results, we discuss the potential of kriging to deal with both linear and nonlinear situations.