[pt] A determinação de parâmetros de interação precisos para equações de estado (EdE) em sistemas aquosos de aminas são cruciais para desenvolver modelos termodinâmicos em processos da engenharia química. O sistema binário de Nmetildietanolamina (MDEA) e água na purificação do biogás foi avaliado usando as abordagens 𝜑 – 𝜑 e 𝛾 – 𝜑, EdE de Peng–Robinson clássico com a regra de mistura não aleatória e EdE Peng–Robinson com a regra de mistura Wong-Sandler, para otimizar o fator acêntrico, ω, de componentes puros, e os parâmetros de interação binária, 𝑘𝑖𝑗. Os parâmetros de interação 𝑢𝑖𝑗 das EdE que incorporam o modelo UNIQUAC, como a abordagem γ – φ e a regra de mistura de Wong-Sandler também foram otimizados. Esses parâmetros foram avaliados usando um algoritmo de pressão de bolha reativa, codificação MATLAB e minimização de funções objetivas relacionadas ao desvio médio absoluto, AAD, entre dados experimentais e calculados em diferentes temperaturas. Os ω calculados de água, CO2 e MDEA foram 0,3275, 0,2039 e 1,0133, respectivamente, com AAD inferior aos valores da literatura. A abordagem 𝜑 − 𝜑 com EdE clássica de Peng–Robinson com regra de mistura Wong– Sandler foi mais adequada para o binário MDEA – H2O, resultando em 𝑢12 0 = −234.2841, 𝑢12 𝑇 = 1.0499, 𝑢21 0 = 266.4326, 𝑢21 𝑇 = 0.1966, 𝑘𝑖𝑗 = −0.0715, com pressão de vapor AAD% = 6,57% e composição AAD% = 17,51%. Devido à natureza altamente não ideal do sistema binário CO2 – H2O, nem as abordagens φ – φ ou γ – φ usando as EdE selecionadas resultaram em diagramas precisos de pressão de ponto de bolha para o equilíbrio vapor – líquido (VLE). / [en] Determining interaction parameter for equations of state (EOS) of water – amines systems are crucial to develop accurate models in chemical engineering processes. The binary system of N-methyldiethanolamine (MDEA) and water in biogas purification was evaluated using both φ – φ and γ – φ approaches, classic Peng–Robinson with the empirical “non-random” mixing rule and Peng–Robinson with the Wong-Sandler mixing rule EOS to optimize pure components acentric factor, ω, and binary interaction parameters, 𝑘𝑖𝑗. The interaction parameters 𝑢𝑖𝑗 from EOS that incorporate UNIQUAC model, such as γ – φ approach and Wong-Sandler mixing rule were optimized as well. These parameters were evaluated using a bubble pressure algorithm, MATLAB coding and minimization of objective functions related to absolute average deviation, AAD, between experimental and calculated data at different temperatures. The calculated ω of water, CO2 and MDEA were 0.3275, 0.2039 and 1.0133 respectively with lower AAD than literature values. The 𝜑−𝜑 approach classic Peng–Robinson with Wong – Sandler mixing rule EOS was better suited for the MDEA – H2O binary, resulting in as 𝑢120=−234.2841, 𝑢12𝑇=1.0499, 𝑢210=266.4326, 𝑢21𝑇=0.1966, 𝑘𝑖𝑗=−0.0715, with vapor pressure AAD% = 6.57% and composition AAD% = 17.51%. Due to the highly non-ideal nature of the CO2 – H2O binary system, neither φ – φ or γ – φ approaches using the selected EOS resulted in accurate vapor – liquid equilibrium (VLE) bubble point pressure diagrams.

This work is focused on thermodynamic modelling of isobaric vapour-liquid-liquid equilibrium (VLLE) (homogeneous) and (heterogeneous) for binary, ternary and quaternary systems. This work uses data for organic/aqueous systems; historically these mixtures were used in the production of penicillin and were required to be separated by continuous fractional distillation. Modelling of the separation required phase equilibrium data to be available so that predictions could be made for equilibrium stage temperatures, vapour compositions, liquid compositions and any phase splitting occurring in the liquid phase. Relevant data became available in the literature and work has been carried out to use relevant theories in correlating and predicting as was originally required in the distillation equilibrium stage modelling. All the modelling carried out was at atmospheric pressure. The modelling has been done using an Equation of State, specifically Peng Robinson Styrjek Vera (PRSV), combined with the activity coefficient model UNIversal QUAsi Chemical (UNIQUAC) through Wong Sandler mixing rules (WSMR). The success of all correlations and predictions was justified by minimizing the value of the Absolute Average Deviation (AAD) as defined within the thesis. Initially the integral Area Method and a method called Tangent Plane Intersection (TPI) were used in the prediction of liquid-liquid equilibrium (LLE) binary systems. This work used a modified 2-point search, suggested a 3-point search and has successfully applied both of these methods to predict VLLE for binary systems. It was discovered through the application of the TPI on ternary VLLE systems that the method was strongly sensitive to initial values. This work suggested and tested a Systematic Initial Generator (SIG) to provide the TPI method with realistic initial values close to the real solution and has demonstrated the viability of the SIG on improving the accuracy of the TPI results for the ternary systems investigated. In parallel with the TPI another method the Tangent Plane Distance Function (TPDF) was also investigated. This method is based on the minimisation of Gibbs free energy function related to the Gibbs energy surface. This method consistently showed it was capable of predicting VLLE for both ternary and quaternary systems as demonstrated throughout this work. The TPDF method was found to be computationally faster and less sensitive to the initial values. Some of the methods investigated in this work were also found to be applicable as phase predictors and it was discovered that the TPDF and the SIG methods were successful in predicting the phase regions; however the TPI method failed in identifying the 2 phase region. Applying the techniques described to newly available quaternary data has identified the strengths and weaknesses of the methods. This work has expanded the existing knowledge and developed a reliable model for design, operation and optimisation of the phase equilibria required for prediction in many separation processes. Currently available modelling simulation packages are variable in their predictions and sometimes yield unsatisfactory predictions. Many of the current uses of VLLE models are particularly focused on Hydrocarbon/Water systems at high pressure. The work described in this thesis has demonstrated that an EOS with suitable mixing rules can model and predict data for polar organic liquids at atmospheric and below atmospheric pressure and offers the advantage of using the same modelling equations for both phases.