Determination of octane number of gasoline by double ANN algorithm combined with multidimensional gas chromatography

In this paper, a double artificial neural network (ANN) algorithm has been established for calculating the octane number (ON) of gasoline from the results of multidimensional gas chromatography analysis. Multidimensional resolution column was applied to obtain the results of the detailed hydrocarbon analysis. The double ANN regression model has been established between the results of the detailed hydrocarbon analysis and the actually determined research octane number (RON). When the method was applied to determine RON of export gasoline samples, the deviation of results was about 0.5 RON compared with the standard method. The result of double ANN regression model was better than the result of partial least square (PLS) regression model. This method was easy to manipulate, and the modelling process was fast and easy to achieve. It was suitable for measuring the ON of the gasoline samples from the refinery and the export inspection.