Necessity of imposing total miscibility for certain binary pairs in LLE data correlations

Abstract

Liquid-liquid equilibrium (LLE) data correlation of multicomponent mixtures is frequently carried out without using any procedure to ensure that the model parameters obtained for the totally miscible binary mixtures are consistent with such behavior (i.e. they do not lead to two liquid phases in equilibrium). In other words, the behavior of the model beyond the LLE region fitted (experimental tie-lines) is not usually considered in LLE correlations. It seems to be wrongly assumed that the mere absence of tie-lines data in totally miscible regions should guarantee that the parameters resulting from the LLE correlation lead to total miscibility in such homogeneous regions. We have checked that for the correlation of certain type 1 ternary systems (Treybal classification) a high probability exists of obtaining inconsistent model parameters, which lead to type 2 instead type 1 ternary behavior, unless specific restrictions for such parameters are imposed during the correlation. Commercial tools to carry out this type of data fitting, which are frequently included in process simulation packages such as Aspen Plus, do not include the possibility of applying such restrictions and consequently they could provide this type of misleading correlation results with serious consequences in equipment design. In the present paper, LLE data correlation for certain type 1 ternary systems has been carried out using Aspen Plus resulting in inconsistent type 2 results. These same systems have been satisfactorily correlated using our own correlation tools including restrictions to ensure the required miscibility behavior, and thus obtaining consistent parameters. A similar problem could occur in the correlation of the less frequent island (or zero) type ternary systems. Some examples are also presented for such type of systems.