Gasoline Brands Recognition And Octane Number Determination Using Near-infrared Spectroscopy

Octane number is one of the most important properties of gasoline, and gasoline is divided into different brands based on octane number. Traditional laboratory analysis is usually used to measure octane number, however, the analytical process is complex, time-consuming with high cost. Near-infrared (NIR) spectroscopy is a non-destructive and fast analysis method, which is preferable to rapidly determine the properties of petroleum products offline or online. This thesis studies the application techniques of NIR spectroscopy and developed a new gasoline octane number analyzer. The main contents of this thesis are as follows:1. Review the development and application of NIR spectroscopy analysis technology; introduce the detailed modeling steps and theory background of NIR calibration model; then present the application of pattern recognition technology in NIR qualitative analysis.2. Apply classification technique in automatic recognition of gasoline brands. Use principal component analysis (PCA) to extract features of spectra, then discriminate gasoline brands by using several classification algorithms, and compare their experimental results. Experiment results show that satisfied classification performance can be obtained by SIMCA (soft independent modeling of class analogy) algorithm, which can be widely applied to the fast recognition of brands of gasoline products.3. hi the NIR calibration modeling process, traditional training sample selection policy only selects the nearest samples by the spectrum distance. This thesis analyzed the shortage of the selection policy and designed an improved training sample selection policy. The improved policy first chooses several nearest training samples, then abandons some abnormal samples by PCA residua, finally build partial least squares (PLS) regression model to predict the octane number of gasoline products. Experiment results show that the new training sample selection policy improves prediction precision of NIR model obviously.4. Based on the above results, a gasoline octane number analyzer using NIR spectroscopy is developed. The analyzer has been applied in a refinery for more than one year. Application results show that the analyzer performs satisfactorily on the prediction precision and stability, which presents very good application values.