Correlations Between Cloud Point and Compositional Properties of Palm Oil and Liquid Fractions from Dry Fractionation. (4th May 2017)

• Record Type: Journal Article
• Authors: Danthine, S.; Lefébure, E.; Blecker, C.; Dijckmans, P.; Gibon, V.
• Is Part Of: Journal of the American Oil Chemists' Society. Volume 94:Number 6(2017)
• Page Start: 841
• Page End: 853
• Publication Date: 2017-05-04
• Subjects: Dry fractionation -- Palm oil liquid fractions -- Cloud point -- Compositional properties -- Statistical analysis; Oils and fats -- Periodicals; Soap -- Periodicals; Fats -- Periodicals; Food-Processing Industry -- Periodicals; Acides gras -- Périodiques; Aliments -- Industrie et commerce -- Périodiques; Huiles et graisses -- Périodiques; Savon -- Périodiques; Oliën; Chemie; Oils and fats; Soap; Periodicals; 547.77
• DOI: 10.1007/s11746-017-2990-2
• Languages: English
• ISSNs: 0003-021X

Abstract

Abstract: Dry fractionation of palm oil can be conducted as a multi‐step process; this gives rise to new softer and harder fractions having multitudes of applications in fat foods. This work focuses on the liquid fractions obtained following a triple‐step fractionation; the cloud point was measured on a set of 125 palm oil, olein, super olein and top olein samples, with the intention to correlate this value to the compositional properties. The Mettler cloud point, the diacylglycerol content, some selected single or groups of triacylglycerols and the iodine value (in total, 11 variables per sample) were measured and entered in a statistical model. A principal component analysis (PCA) was first carried out from which different sub‐groups were highlighted. In each sub‐group, various regressions (simple linear, simple non‐linear and multiple linear) were applied and 28 significant equations were derived. Out of these, one multiple linear regression involving the iodine value, the UUU and PPP contents showed the best correlation with the Mettler cloud point. This equation was selected to predict the cloud point since it was further successfully validated by using a set of 25 other independent samples. A partial least square (PLS) regression was tested and also considered adequate to predict the cloud point.