Rapid and accurate determination of diesel multiple properties through NIR data analysis assisted by machine learning

Area of Science:

• Petrochemical industry
• Analytical chemistry
• Spectroscopy

Background:

• Accurate diesel property detection is crucial for quality assessment and environmental protection.
• Current methods for diesel quality analysis are often inefficient and costly.

Purpose of the Study:

• To develop a novel machine learning model for rapid and accurate simultaneous determination of multiple diesel properties using near-infrared (NIR) spectroscopy.
• To improve the efficiency and reduce the cost of diesel quality detection.

Main Methods:

• Development of a hybrid machine learning model combining improved XY co-occurrence distance (ISPXY) and differential evolution-gray wolf optimization support vector machine (DEGWO-SVM).
• Application of the model to near-infrared (NIR) spectroscopy for analyzing diesel samples.

Main Results:

• The developed model achieved superior performance in terms of average recovery, mean square error, mean absolute percentage error, and determination coefficient compared to existing methods.
• Demonstrated the capability for simultaneous determination of diesel density, viscosity, freezing point, boiling point, cetane number, and total aromatics.

Conclusions:

• The proposed hybrid model offers a significant advancement for efficient and cost-effective diesel quality detection.
• The model shows strong potential for routine diesel monitoring applications in the petrochemical industry.