Avaliação de modelos termodinâmicos para caracterizar o sistema benzeno-tiofeno

Abstract

The separation of benzene and thiophene is a relevant challenge in chemical engineering due to the low relative volatility difference between the components and the presence of sulfur in thiophene, which compromises fuel quality and catalytic activity in refining processes, making it essential to employ thermodynamic models capable of accurately representing the vapor–liquid equilibrium (VLE). In this work, different models available in the Aspen Plus simulator were evaluated to represent the binary benzene–thiophene system, using experimental data available in the literature as reference. The qualitative analysis consisted of comparing simulated T–x–y curves with experimental points, while the quantitative analysis was carried out through the calculation of root mean square deviations (RMSD) in temperature and vapor-phase composition. The results showed contrasting performances: the Peng–Robinson equation of state presented the lowest deviation in temperature (RMSD = 0.1845 °C), confirming its robustness in thermal prediction, while the Soave–Redlich–Kwong model stood out in representing the vapor phase, with RMSD in composition of only 0.0009, evidencing greater fidelity in predicting equilibrium compositions. The UNIQUAC model demonstrated balanced performance, with RMSD of 0.2364 °C in temperature and 0.0017 in composition, followed by NRTL, which presented 0.3448 °C and 0.0032, respectively, confirming intermediate results. In contrast, the correlational models Chao–Seader and Grayson–Streed exhibited the largest deviations, with RMSD in temperature of 1.0735 °C and 1.2690 °C and composition errors above 0.008, while UNIFAC, despite its predictive nature, showed RMSD of 0.8707 °C in temperature and 0.0090 in composition, confirming its limitations for aromatic systems containing sulfur. Therefore, it is concluded that Peng–Robinson and SRK are the most suitable models to represent the studied system, each excelling in a specific variable, with UNIQUAC emerging as an intermediate alternative capable of providing consistent results in both variables, although without surpassing the performance of cubic equations of state.