Determination of biodiesel content in biodiesel/diesel blends using NIR and visible spectroscopy with variable...

Area of Science:

• Analytical Chemistry
• Spectroscopy
• Chemometrics

Background:

• Accurate quantification of biodiesel content in biodiesel/diesel blends is crucial for quality control and regulatory compliance.
• Visible and near-infrared (NIR) spectroscopy offers a potential rapid and non-destructive analytical approach.

Purpose of the Study:

• To evaluate the effectiveness of visible and NIR spectral ranges, individually and combined, for determining biodiesel content in biodiesel/diesel blends.
• To compare chemometric models, including Multiple Linear Regression (MLR), Partial Least Squares (PLS), and variable selection techniques like Successive Projections Algorithm (SPA), Stepwise (SW), and Jack-Knife (Jk).

Main Methods:

• Spectroscopic analysis was performed on 100 biodiesel/diesel blends (5-50% v/v biodiesel).
• Data preprocessing involved Savitzky-Golay derivative filtering (2nd order, 17-point window) with offset correction.
• Chemometric models (MLR, PLS) were developed and optimized using variable selection algorithms (SPA, SW, Jk).

Main Results:

• The Successive Projections Algorithm coupled with Multiple Linear Regression (SPA-MLR) achieved the best performance in the NIR and visible regions, utilizing only two and eight wavelengths, respectively, with low RMSEP values (0.6439% and 0.5741%).
• In the combined visible-NIR region, the Stepwise (SW)-MLR model using five wavelengths yielded an RMSEP of 0.9533%.
• Variable selection significantly reduced the number of wavelengths required and improved prediction accuracy.

Conclusions:

• Both visible and NIR spectral ranges are suitable for developing rapid, non-destructive methods to quantify biodiesel in mineral diesel blends.
• Variable selection techniques are highly effective in enhancing the performance and simplifying the spectral models for biodiesel quantification.