Rapid Determination of Palmitic Acid Content in Edible Oils Using Vis-NIR Reflectance Spectroscopy and Deep Learning...

Area of Science:

• Analytical Chemistry
• Food Science
• Spectroscopy

Background:

• Fatty acid composition is crucial for edible oil quality assessment.
• Accurate quantification of specific fatty acids like palmitic acid is essential.
• Existing methods can be time-consuming and destructive.

Purpose of the Study:

• To develop a rapid and non-destructive method for predicting palmitic acid content in edible oils.
• To evaluate the effectiveness of visible-near-infrared (Vis-NIR) reflectance spectroscopy combined with deep learning.
• To compare the performance of various machine learning and deep learning models.

Main Methods:

• Collected 1740 Vis-NIR reflectance spectra (350-2500 nm) from 87 edible oil brands.
• Determined reference palmitic acid values using gas chromatography-mass spectrometry (GC-MS).
• Developed and compared SVR, KNN, 1D-CNN, 1D-ResNet, 1D-Inception, and 1D-Inception-ResNet models using full spectra and CARS-selected wavelengths.

Main Results:

• The 1D-ResNet model achieved Rp2=0.9027 and RMSEp=1.13 with full-spectrum data.
• The 1D-Inception-ResNet model, using 91 CARS-selected wavelengths, achieved superior results with Rp2=0.9825 and RMSEp=0.4804.
• Deep learning models significantly outperformed conventional methods.

Conclusions:

• Vis-NIR reflectance spectroscopy coupled with informative wavelength selection and deep learning offers an effective strategy for rapid palmitic acid prediction in edible oils.
• This approach enables efficient and non-destructive quality control of edible oils.
• The study highlights the potential of advanced spectral analysis and AI for food analysis.