[Study on Refined Oil Identification and Measurement Based on the Extension Neural Network Pattern Recognition]

Area of Science:

• Analytical Chemistry
• Environmental Science
• Chemical Engineering

Background:

• Fuel consumption significantly contributes to air pollution, necessitating urgent solutions for large consumption, low quality, and emission control.
• Accurate identification and measurement of refined oil products are vital for effective air pollution monitoring and control strategies.
• Existing methods for fuel analysis may lack the speed and accuracy required for real-time monitoring and regulatory compliance.

Purpose of the Study:

• To develop a rapid and accurate method for identifying and quantifying refined oil products, including gasoline, diesel, and kerosene.
• To improve the efficiency of network model identification for refined oil characterization.
• To address the challenges of "over-fitting" in classification models through advanced techniques.

Main Methods:

• Principal Component Analysis (PCA) was employed for data dimension reduction and to extract finer characteristic parameters from three-dimensional fluorescence spectra (Excitation-Emission Matrix, EEM).
• A neural network model was designed for both qualitative identification and quantitative measurement of refined oil types.
• Cross-validation was utilized during classification to prevent the "over-fitting" phenomenon.

Main Results:

• The developed method achieved an average recognition rate of 0.99 for refined oil products.
• The average recovery rate for concentration measurement reached 0.95.
• The average pattern recognition time was 2.5 seconds, significantly faster (48.5% reduction) than the PARAFAC model analysis method.

Conclusions:

• The combined approach of PCA and neural networks offers a highly effective and efficient solution for the rapid and accurate identification and measurement of refined oil products.
• This method significantly enhances operational speed and demonstrates ideal application effects for air pollution monitoring.
• Accurate analysis of complex mixtures like refined oil requires specific calibration samples to ensure precision and accuracy.