In the production of petroleum products, the study of IR absorption spectra is most often used to analyze the properties of a mixture. The priority of this method is due to the fact that the characteristics of the IR spectrum are directly related to the nature (structure and chemical composition) of the absorbing substance, and also depend on the aggregate state of the substance, temperature, pressure, etc. The unambiguity of the relationship between the molecular structure of a substance and its IR spectrum allows us to determine the composition of the mixture. For this purpose, calibration models should be built that connect the IR spectrum with the value of the quality indicators of petroleum products. The paper considers the methods of creating calibration models proposed by various authors, presented in well-known literary sources. To create calibration models in this paper, it is proposed to use the method of principal components and neural network modeling. Also, in order to increase the reliability of the automated control system for compounding motor fuels, it is proposed to use virtual analyzers (VA) of quality indicators of communication models, the quality indicator calculated from the calibration models of the flow analyzer with the corresponding technological variables of the compounding process. The output of the calibration models is also used for adjusting the VA.
Keywords: Oil refining, IR spectrometer, quality indicators of petroleum products, flow IR quality analyzer, calibration models of the analyzer, virtual quality analyzer