Authentication of edible oils using an infrared spectral library and digital sample sets: A feasibility study

• Published in: Journal of chemometrics 2023-03, Vol.37 (3), p.n/a
• Main Authors: Sota‐Uba, Isio, White, Collin G., Booksh, Karl, Lavine, Barry K.
• Format: Article
• Language: English
• Subjects: Adulterants; adulteration; Binary mixtures; Classification; Datasets; Detection limits; Edible oils; Experimental data; Feasibility studies; Fourier transforms; Genetic algorithms; Infrared radiation; Infrared spectra; Infrared spectroscopy; Libraries; mixture analysis; Oils & fats; Olive oil; Pattern recognition; Spectroscopy; Spectrum analysis; variable selection
• ISSN: 0886-9383; 1099-128X
• DOI: 10.1002/cem.3469

A potential method to determine whether two varieties of edible oils can be differentiated by Fourier transform infrared (FTIR) spectroscopy is proposed using digitally generated data of adulterated edible oils from an infrared (IR) spectral library. The first step is the evaluation of digitally blended data sets. Specifically, IR spectra of adulterated edible oils are computed from digitally blending experimental data of the IR spectra of an edible oil and the corresponding adulterant using the appropriate mixing coefficients to achieve the desired level of adulteration. To determine whether two edible oils can be differentiated by FTIR spectroscopy, pure IR spectra of the two edible oils are compared with IR spectra of two edible oils digitally mixed using a genetic algorithm for pattern recognition to solve a ternary classification problem. If the IR spectra of the two edible oils and their binary mixtures are differentiable from principal component plots of the spectral data, then differences between the IR spectra of these two edible oils are of sufficient magnitude to ensure that a reliable classification by FTIR spectroscopy can be obtained. Using this approach, the feasibility of authenticating edible oils such as extra virgin olive oil (EVOO) directly from library spectra is demonstrated. For this study, both digital and experimental data are combined to generate training and validation data sets to assess detection limits in FTIR spectroscopy for the adulterants. A potential method to determine whether two different varieties of edible oils can be differentiated by Fourier transform infrared spectroscopy is proposed using digitally generated data of adulterated edible oils from an infrared spectral library that contains the infrared spectra of 20 different varieties of plant‐based edible oils collected over a 3‐year period encompassing multiple brands representing supplier to supplier variation and seasonal and batch variation within a supplier.