Quality control of fragrances using Raman spectroscopy and multivariate analysis

An analytical methodology using Raman spectroscopy and chemometrics was developed for direct, fast and non‐destructive discrimination and prediction of the properties of fragrances according to their composition. The soft independent modeling of class analogies was used as a supervised classification method for fragrances classification, and partial least squares regression as a multivariate calibration method for the prediction of physicochemical properties of fragrances, such as density and refractive index. From 155 fragrance samples, the model exhibited a high success rate for all of the studied fragrance classes, with 100% correct classification. In the multivariate calibration model, adequate correlation was observed between the measured and partial least squares regression‐predicted data for refractive index and density, with a relative standard error of prediction between 0.02% and 0.07%, respectively. This study demonstrates the wide applicability of the methodology for the discrimination, classification, and prediction of complex olfactory mixtures in quality control of fragrances. Copyright © 2015 John Wiley & Sons, Ltd. An analytical methodology using Raman spectroscopy and chemometrics was developed for direct fast and non‐destructive discrimination and prediction of the properties of fragrances according to their composition. The samples used to validate the model exhibited a high success rate for all of studied fragrance classes, with 100% correct classification. In the multivariate calibration model, adequate correlation was observed between the measured and PLS‐predicted data for refractive index and density, with a relative standard error of prediction between 0.02% and 0.07%, respectively.