Calculation of liquid Cp of pure compounds using an improved thermodynamic model based on group contributions and artificial neural networks.

• A new model for heat capacity calculation of pure compounds in liquid phase was developed. • This model is based on artificial neural networks and group contribution theory. • This new model outperformed the group contribution model of Aspen PlusⓇ simulator. Liquid Cp of pure compounds is essential for process modeling, optimization, and design. However, the experimental information reported for the heat capacities of new chemical and emerging compounds is limited, and the predictive models that have been reported for the estimation of this property may show high modeling errors, mainly in specific regions of the phase diagram. This paper reports the development, testing, and comparison of an improved thermodynamic model for the calculation of this property by exploiting the synergy between group contributions and artificial neural networks. An extensive database constituted by 22715 experimental values of liquid heat capacity ranging from 6.71E+04 to 1.11E+06 J/kmol · K over a wide temperature range (85.59–860 K) of 441 compounds was used for the model development. Different input variables and artificial neural networks configurations were evaluated to select the best thermodynamic model for calculating this thermodynamic property. The performance of this improved model was characterized in terms of its capabilities and limitations for Cp calculation of different chemical families where the impact of the molecular weight and reduced temperature was also discussed. The results showed that the thermodynamic model incorporating reduced temperature as an input parameter improved the Cp calculations near to the critical point. The calculation errors of this new model ranged from 0.47 to 7.72% for most of the analyzed chemical families. This new model was compared with the Cp model used by the Aspen PlusⓇ process simulator and the results showed its superior capabilities and estimations. Therefore, it can be easily implemented in commercial process simulators for Cp calculation of different chemical families, over a broad temperature range, as an alternative reliable thermodynamic tool for process systems engineering. [ABSTRACT FROM AUTHOR]