In the present research we discuss a novel way to set up a predictive method for determining the water content of oil based on the Partial Least Squares (PLS) regression analysis of reflectometric signals. Ten different extra virgin olive oils with a water content ranging from 714 to 2008 mg of water/kg of oil were submitted to reflectometric measurement by means of a hand made probe connected to a digital sample oscilloscope with TDR functions interfaced with a PC. Limits of the classical approach based on the TDR electromagnetic theory in the prediction of these small water content differences were also discussed. The results show that the suggested novel approach is able to predict the water content of very small quantities of oil (1.8 g) in a 3 ml translucent disposable PE cuvette by means of PLS regressions characterized by R^2 value up to 0.984 and a root mean square error of prediction of about 55 mg of water/kg of oil. The temporal region showing the best information content corresponded with the rise of the reflection of the probe end, but information highly correlated with the water content can be extracted from other temporal regions of the entire TDR signal.
Assessment of the water content in extra virgin olive oils by Time Domain Reflectometry (TDR) and Partial Least Squares (PLS) regression methods
RAGNI, LUIGI;BERARDINELLI, ANNACHIARA;CEVOLI, CHIARA;VALLI, ENRICO
2012
Abstract
In the present research we discuss a novel way to set up a predictive method for determining the water content of oil based on the Partial Least Squares (PLS) regression analysis of reflectometric signals. Ten different extra virgin olive oils with a water content ranging from 714 to 2008 mg of water/kg of oil were submitted to reflectometric measurement by means of a hand made probe connected to a digital sample oscilloscope with TDR functions interfaced with a PC. Limits of the classical approach based on the TDR electromagnetic theory in the prediction of these small water content differences were also discussed. The results show that the suggested novel approach is able to predict the water content of very small quantities of oil (1.8 g) in a 3 ml translucent disposable PE cuvette by means of PLS regressions characterized by R^2 value up to 0.984 and a root mean square error of prediction of about 55 mg of water/kg of oil. The temporal region showing the best information content corresponded with the rise of the reflection of the probe end, but information highly correlated with the water content can be extracted from other temporal regions of the entire TDR signal.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
