[A novel fuzzy neural network method for diesel quantitative analysis with near infrared spectroscopy]

To satisfy the requirement of the production control and quality inspection of petrochemical products, a novel fuzzy neural network control method is proposed for determining product composition by near infrared spectroscopy. For the data analysis three different diesel products were selected as samples and six analytical models, such as saturated hydrocarbons, polar compounds, monoaromatics, dicyclic aromatics, tricyclic aromatics and naphthenes, were developed with the proposed fuzzy neural network method. Based on dSPACE, the near infrared spectroscopy system real-time experimental platform has been established to testify and analyze different diesel samples. The experimental results show that the improved performance is superior because of its advantage of quick response and good robustness. The mean squared error (MSE) of calibration and prediction samples is of the order of 10(-6) in the spectral range of 800-2300 nm. The developed method can be used in the research on petrochemical products processing.