[Determination of Gasoline Composition Based on Raman Spectroscopy]

Area of Science:

• Analytical Chemistry
• Spectroscopy
• Chemometrics

Context:

• Accurate and rapid determination of gasoline composition is crucial for quality control and regulatory compliance.
• Traditional methods like gas chromatography can be time-consuming and require extensive sample preparation.
• Developing faster, non-destructive analytical techniques for gasoline analysis is an ongoing research area.

Purpose:

• To develop and validate a rapid, precise, and convenient method for predicting the composition of gasoline using Raman spectra.
• To compare the performance of multi-output least squares support vector regression (LS-SVR) with partial least squares combined with multiple regression analysis (PLS-MRA) for gasoline component prediction.

Summary:

• Raman spectra of 410 gasoline mixture samples (brands 93 and 97) with varying aromatic, olefin, benzene, methanol, and ethanol content were collected and preprocessed.
• Characteristic spectral peaks were extracted as eigenvalues, and two chemometric models (multi-output LS-SVR and PLS-MRA) were built using these eigenvalues.
• The multi-output LS-SVR model demonstrated superior performance, achieving low root mean square errors (e.g., 0.14% for ethanol) and high correlation coefficients (e.g., 0.9993 for aromatic content).

Impact:

• The developed Raman spectroscopy-based LS-SVR method provides a precise, fast, and convenient alternative for gasoline composition detection.
• This technique can be effectively applied to the quality control of gasoline throughout its production, transportation, and storage.
• The model's accuracy, with prediction errors not exceeding 0.5% for unknown samples, meets industrial measurement requirements.