Edible oils are mixtures of triglycerides that differ in their composition of fatty acids. Analysis of edible oils to detect adulteration, to authenticate, and to classify by type was demonstrated by Raman spectroscopy approximately twenty years ago.  More recently, multivariate classification and calibration methods have been applied to Raman and IR spectra of edible oils to improve the classification rate, lower the detection limits for adulterants, and quantitatively assess product quality.  Generally, classification rates for edible oils are around 90% and detection limits are around 10% as reported in recently published work.  A Raman data set consisting of 297 spectra spanning 15 oil types was collected using a 785nm laser, Ocean Optics portable spectrometer and an InPhotonics probe fitted with a custom made sapphire ball lens dip probe.  Spectra were collected at multiple integration times to test the robustness of the target methods to different noise levels. Edible oil samples were collected from different companies and vendors (same company, different batches and possible different manufacturing plants).  This was necessary to build better statistical distributions of expected in-class variance to determine target method performance when developing classification models from training sets and to have sufficient samples to construct independent training and validation sets. Using a genetic algorithm for spectral pattern recognition, we demonstrate that sample variability is a major source of variation comparable to edible oil type in the Raman spectral data. The significance of this result is discussed in the context of authentication and the detection of adulterants in edible oils.<br/>

Authentication of edible food oils using raman spectroscopy

256th National Meeting (2018)

The official archive and repository for all ACS Meetings & Expositions abstracts and select posters or presentations.