Effect of Variation between Different Experimental VLE Data Sets on Thermodynamic Model and Separation Predictions: NRTL Correlation of the Ethanol plus Water System

The effect of correlating the NRTL model against various ethanol + water VLE data sets available in the NIST ThermoData Engine database on the resultant model binary interaction parameters and the subsequent separation prediction is investigated. To mimic industrial application, the study was conducted in Aspen Plus V11.0 using the default convergence criteria and algorithms. It was found that model parameters correlated to the different data sets as well as the separation prediction have the same magnitude of variation as propagated experimental uncertainty in the data. Further, the same phase behavior and separation prediction are obtained if a particular data set is used for the correlation, irrespective of which model parameters are correlated. For the process simulation, it was found that the effect in variation of the model parameters is the greatest in the top and bottom composition; however, the variation was never greater than 0.03 mole fraction. It was also found that more variation is present for separation sequences operating close to a pinch point, e.g., close to the minimum reflux ratio. Finally, for the ethanol + water system, the default parameters in Aspen Plus V11.0 appear to result in a good prediction of the phase behavior of the system. Further studies should extend this work to other systems and other thermodynamic models.