Application of Multivariate Analysis in Mid-Infrared Spectroscopy as a Tool for the Evaluation of Waste Frying Oil Blends

Abstract

Mid-infrared spectroscopy, in association with multivariate chemometric techniques, was employed for pattern recognition and the determination of the composition of waste frying oils (WFO); data are presented in terms of the percentage of soybean oil, palm oil and hydrogenated vegetable fat in frying oil blends. Principal component analysis (PCA) was performed using spectral data (3,000–600 cm−1) to discriminate between the samples containing 100% soybean oil, 100% palm oil, 100% hydrogenated vegetable fat groups and their blends. Additionally, the results indicated that partial least squares (PLS) models based on mid-infrared spectra were suitable as practical analytical methods for predicting the oil contents in WFO blends. PLS models were validated by a representative prediction set, and the root mean square errors of prediction (RMSEP) were 2.8, 4.7 and 5.5% for palm oil, soybean oil and hydrogenated vegetable fat, respectively. The proposed methodology can be very useful for the rapid and low cost determination of waste frying oil composition while also aiding in decisions regarding the management of oil pretreatment and production routes for biodiesel production.