A rigorous method to evaluate the consistency of experimental data in phase equilibria. Application to VLE and VLLE

Abstract

This work forms part of a broader study that describes a methodology to validate experimental data of phase equilibria for multicomponent systems from a thermodynamic-mathematical perspective. The goal of this article is to present and justify this method and to study its application to vapor–liquid equilibria (VLE) and vapor–liquid–liquid equilibria (VLLE), obtained under isobaric/isothermal conditions. A procedure based on the Gibbs-Duhem equation is established which presents two independent calculation paths for its resolution: (a) an integral method and (b) a differential method. Functions are generated for both cases that establish the verification or consistency of data, δψ for the integral test and δζ for the differential approach, which are statistically evaluated by their corresponding average values [δψ, δζ], and the standard deviations [s(δψ), s(δζ)]. The evaluation of these parameters for application to real cases is carried out using a set of hypothetical systems (with data generated artificially), for which the values are adequately changed to determine their influence on the method. In this way, the requirements of the proposed method for the data are evaluated and their behavior in response to any disruption in the canonical variables (p,T, phase compositions). The conditions for thermodynamic consistency of data are:δψ<2, s(δψ)<0.2δζ<5. In systems with VLLE, in addition to the previous criteria, must occur that: δXLLE<0.05 and δTLLE<0.5. The new proposed method has been tested with a set of 300 experimental binary systems, biphasic and triphasic, obtained from published bibliography, and the results are compared with those of other tests commonly used for testing thermodynamic consistency. The results show that the greater rigor of the proposed method is mainly due to the simultaneous verification of various independent variables. As a result, the conditions for the new test are verified for fewer systems than using other tests mentioned in the literature (i.e., Fredenslund-test and direct of Van Ness). Its unique application is sufficient to ensure the consistency of experimental data, without using other tests.