Análise estatística de modelos termodinâmicos baseados em NRTL para eletrólitos

Abstract

Electrolytic solutions have great application in industrial processes so rigorous and accurate calculation of thermodynamic properties of electrolytes consists of a solid basis for phase equilibrium calculations and also for a simulation of chemicall processes withinelectrolytes. Therefore,a validation of the model and its parameters is interesting and important to have confidence in the prediction.Thus, two Gibb s energy models in excess, e-NRTL and refined e-NRTL are qualitatively and statistically analyzed in this work. The preliminary analysis with the case of study aims to evaluate the behavioras well as the suitability of the model (Fisher s test) in electrolyte systems of different complexities: aqueous electrolyte system, aqueous multi-electrolyte system and mixed solvent electrolyte system.The case of study are done at a temperature of 298.15 K, except the mixed solvent electrolyte system which is done at 298 K. Subsequently, the statistical evaluation of the models is done, in each type of electrolytic system, using the parametric identification through local sensitivity analysis based on Gram-Schmidt orthogonalization.It is noteworthy that in parametric identifiability, the temperature is considered a factor for the evaluation of aqueous electrolyte systems and aqueous multielectrolyte system, whereas,in the evaluation of mixed solvent electrolyte system, the temperature is kept constant at 298 K. In addition, the present work uses binary parameters, both in the case study and in the statistical evaluation, previously estimated and obtained in the literature.We also investigate the influence of the identifiable parametric set in the parametric confidence region. Therefore, the determination of the parametric confidence regionof both thermodynamic models is proposed from two set of identifiable parameters being the first set corresponding to all the selected parameters, while the second set excludes the identifiable parameter of smaller effect in the prediction.In the present work, it is ratified that both thermodynamic models are able to represent qualitatively the behavior of electrolytic systems, however system with high electrolyte concentration and/or high amount of non-aqueous solvent are not well represented.From statistical considerations, the suitability of the models to the experimental data reported in the literature is performed by determining the confidence interval of the proportionality constant (β), referring to the experimental variance, which satisfies the equivalence of variances according to Fisher's test.Thus, low compatible experimental errors with modeling in aqueous electrolyte systems and multielectrolyte systems are observed.In contrast, high experimental errors are compatible with electrolytic modeling with mixed solvent, mainly due to the experimental set.Parametric identification only selects parameters from the short-range contribution term, so the usual distance of closest approach (a) is adequate.It is verified that both models over-dimension the parametric confidence regions of the electrolytic system, except the mixed solvent electrolyte system for which open regions are obtained. Furthermore, it is shown that confidence regions are influenced by the set of estimable parameters