Detection of defective virgin olive oils by metal-oxide sensors

An array of metal-oxide sensors was set up for detecting the vinegary, rancid, and fusty defects in virgin olive oils. Spanish Farga and Picual olive oil varieties were spiked with these defective sensory attributes. The optimization process was carried out taking into account the variables affecting the process that were evaluated by three desirability functions. Repeatability studies for six months and within-day were done to evaluate the sensor responses. The responses were preprocessed applying four selected zones (windowed time slicing) of the sensor responses. Assessors carried out the sensory evaluation while the quantification of volatiles was made by gas chromatography (dynamic headspace and thermal desorption). Canonical correlation was applied to study sensor responses versus sensory evaluations and volatile concentrations. Total phenols and fatty acids were quantified for a better understanding of rancid process induced by ultraviolet beam. Fusty attribute was explained by sensor responses (Multiple-R=0.98) using multiple linear regression on the principal components. Samples were completely distinguished according to their defective attributes by applying stepwise linear discriminant analysis on first four factors of PCA.