Optimal management of oil content variability in olive mill batches by NIR spectroscopy

Abstract

Total oil content (OC) is one of the main parameters used to characterize the whole of olives entering a commercial mill, quantified by the total fresh weight of the lot and the oil concentration (%) assessed in a representative sample on olive paste, by means of chemical extraction. Nuclear magnetic resonance (NMR) and NIR spectroscopy are alternative methods even at individual olives. This work evaluates several strategies to calibrate precise NIR models for the estimation of the total OC. To this end, 278 olives were analysed covering whole season variability in terms of olive fresh-weight and the corresponding OC by chemical extraction in 31 batches. The average spectra from hyperspectral NIR images (1003–2208 nm) were computed for each fruit and the actual OC (g) of those olives determined by NMR (0.09 to 1.29 g with a precision of 0.017 g). According to the results, current batch based assessment of the OC (Soxhlet, %) in mills only reproduces 44% of the underlying heterogeneity, despite being the factory standard. The incorporation of individual NIR spectra (278) to the 31 Soxhlet values of the batches allows a 67% explanation of the OC (%) of olives. When estimating OC (g) gathering individual fresh weight and the estimation of oil concentration in olives, a standard error of prediction of 0.061 g is reached (r2 = 0.93), a precision value that approaches the potential limit according to the NMR reference (0.017 g).

Rapid adulteration detection of cold pressed oils with their refined versions by UV–Vis spectroscopy

“…Some physico-chemical parameters, such are: density, acid index, peroxide value, viscosity, or GC–MS (gas chromatography mass spectrometry) detection may be used to distinguish a cold pressed oil from a refined one 3 . Fatty acids composition of vegetable oils 5 – 9 , influences the human health, lipids being among fundamental nutrients 10 .…”

A Model Based on Clusters of Similar Color and NIR to Estimate Oil Content of Single Olives

“…The colors of the olives have been classified using models based on vector supports for the analysis of their image [2]. Currently, color has not stopped being used to build models for estimating oil content (OC), as the first classification criterion although color does not sufficiently explain OC, whereas the NIR spectrum can [3,4].…”

“…NIR technology has a high penetration of laboratory use and oil press as a tool for estimation OC in olive paste, but not in fresh olives, because the low sample variability estimates the performance variability of the samples lower the estimation performance [5]. The management of these variables has been approached by applying Partial Least Square regressions (PLS) or principal components regressions, allowing for the control of the variables that affect the OC, such as the number of samples and maturity of the olives [6], variety [7], fresh weight [3], olive size and year [5], among others. Another problem with the OC estimation models in individual olives is that the heteroscedasticity assumption is frequently breached, that is, the variance of errors is not regular in all the observations of individual olives, ultimately threatening the performance of the model.…”

“…Another problem with the OC estimation models in individual olives is that the heteroscedasticity assumption is frequently breached, that is, the variance of errors is not regular in all the observations of individual olives, ultimately threatening the performance of the model. To optimize the management of variability in the OC estimation, a progressive clustering methodology has been proposed by similar color, NIR spectrum and OC subsequently, improving significantly the prediction error of the model [3].…”

“…The official measurement of the OC using traditional Sohlext methodology [8] needs more than a single olive to be properly determined, necessarily requiring a batch of them, regardless of the individual characteristics of each olive in the group; before taking the oil measurement, the individuals are ground and mixed until the sample is homogenized [8]. However, when one wants to know the OC of individual olives, the Nuclear Magnetic Resonance (NMR) and/or the micro-Sohlext method give good results [3,5,9]. However, most oil mills do not have this type of microanalysis, let alone NMR, due to its high human and technical resource requirement.…”