On the discrimination of models for the viscometric properties of myristic, palmitic and stearic acids and their binary mixtures

Abstract

Fatty acids have a great importance in several industrial activities like the manufacture of ingredients for foods, paints, soap, detergents and a wide array of other products in the fast moving consumer goods industry. Processing of fatty acids requires a good knowledge of transport properties, such as viscosity. This work presents a compilation of experimental values of the viscosity of three long-chain saturated fatty acids, namely: myristic, palmitic and stearic acid. First, a study on the dependence of the viscosity of these fatty acids with temperature is presented comparing results with previously existing data in literature and validating data and models statistically. Then, a thorough analysis of the viscosity of binary mixtures of the aforementioned fatty acids was conducted. To do so, such property was measured over the entire range of compositions in increments of the molar fraction of the components of 0.1 at temperatures ranging from above the melting points of the fatty acids to 373.15 K and correlation with respect to composition and temperature with the Jouyban-Acree model. Last, the so-called viscosity deviations were computed and correlated with Redlich-Kister equations to explain the variation as a function of the molar composition of the corresponding mixture. Discrimination of different models to correlate the experimental data was made on the basis of different statistical criteria.

Practical applications: Good knowledge of transport properties like viscosity is important to understand the physics and processing of materials. Contrary to simply using a viscometer, here cone and plate rheometry is used to accurately determine the viscosity of myristic, palmitic and stearic acids and their binary mixtures, which have high melting points compared to other liquids. In addition, the presentation of statistical discrimination tools like information criteria are highly relevant to understand the goodness of fit of models and the selection of the most statistically relevant equations without using more parameters than strictly necessary.

http://ift.tt/2AYlQw0