Effect of Phase-Equilibrium Uncertainties on the Separation of Heterogeneous Azeotropes—Application to the Water + 1-Butanol System

Abstract

This paper extends application of the Margules-based phase-equilibrium uncertainty method to a system exhibiting a heterogeneous azeotrope. The author developed the method in order to provide practicing engineers with an intuitive and easy-to-apply procedure to quantitatively relate process-design uncertainties to uncertainties in correlated physical properties, specifically nonideal phase equilibrium. The methodology was first applied to two case studies(1) a propylene–propane superfractionator for which small changes in correlated relative volatilities have a large effect on the design of the distillation column; and (2) a dehexanizer column that separates a mixture containing many close-boiling hydrocarbon componentsand demonstrated that the proposed method provides quantitative insight into the effect of property uncertainties for both these diverse process designs and helps to quantify the safety factors that need to be imposed upon the design. In a subsequent study, the methodology was applied to a distillation train that separates the acetone–chloroform–benzene ternary mixture, which contains one maximum-boiling azeotrope, and showed that the approach quantifies the effect of property uncertainties on utility consumption and also identifies limits on operating variables (specifically minimum recycle flow). In this paper, the Margules perturbation method is applied to the separation of the water + 1-butanol binary mixture. The application is complex because this binary exhibits liquid–liquid equilibrium and forms a minimum-boiling azeotrope, and the typical separation scheme includes a decanter as well as two distillation columns. Further, the chosen activity-coefficient model is unable to correlate the data within measurement uncertainty. Nevertheless, the methodology is demonstrated to provide useful quantitative estimates of the design uncertainty resulting from the combined measurement/model phase-equilibrium uncertainty.

Effect of Phase-Equilibrium Uncertainties on Ethyl Acetate Purification

“…The perturbation scheme for the mixture nonideality has been published by Mathias ,, and hence is only briefly presented here. The approach uses the approximation that the perturbed activity coefficients can be shaped by the Margules equation: where γ i is the activity coefficient of component i after perturbation, γ i m is the activity coefficient calculated by the chosen model (the “best” model in the region of interest should be used), and γ i p represents the perturbation obtained through the estimated model uncertainty.…”

“…As in the previous case studies by Mathias, ,, the perturbation scheme has been implemented as a user activity-coefficient model in the NRTL-RK property option in Aspen Plus V8.8. The NRTL-RK property option uses the NRTL activity-coefficient model and the Redlich–Kwong equation of state for the vapor phase.…”

“…Similar to the work of Mathias, for ternary VLE data the percent differences between the model and experimental data for the component K values are used as the measure of combined model uncertainty. Figures – present comparisons between the model K values and the data of Mertl and Lee, Chen, and Huang .…”

“…The exercise showed that phase-equilibrium uncertainties have less effect on synthesis (e.g., flowsheet structure) than on column performance (e.g., utility consumption) and therefore that phase-equilibrium uncertainties may be approximated or even ignored in process synthesis but must be quantified in process design. In a third study, the methodology was applied to the separation of the water + 1-butanol binary mixture, which is a part of processes to purify butanol produced by fermentation . The separation is complex because the binary mixture exhibits liquid–liquid equilibrium and forms a heterogeneous azeotrope and the typical separation scheme includes a decanter as well as two distillation columns.…”

“…The methodology used here follows that of the previous three studies. ,, The first step is comprehensive data regression to obtain the best model fit in the region of application (atmospheric pressure, about 101.3 kPa) and quantitative estimation of the correlation uncertainties. This is a necessary first step.…”