Facile and Accurate Calculation of the Density of Amino Acid Salt Solutions: A Simple and General Correlation vs Artificial Neural Networks

The extensive application of amino acid-based solvents has led to a growing demand for their thermophysical properties. A breakthrough in green carbon capture is anticipated by amino acid salt solutions (AASs), the properties of which should be calculated beforehand. This study develops facile and accurate models for the density calculation of AASs, which would be useful in further process simulations. A general, simple, and easy-to-use correlation for the densities, which is capable of predicting a vast variety of AASs precisely, is first developed. Then, artificial neural networks (ANNs) are assessed to train and test these systems. The developed correlation and ANNs are based on an extensive density databank (2007 data points of 28 AASs) that was prepared from various amino acid and alkaline compounds at different solution concentrations and temperatures. Both calculation procedures take into account the impact of temperature, concentration (wt %), and the molecular weights of amino acid and alkaline compounds as the input variables. To develop the correlation, 68% of the collected data was employed in model training, while 32% was utilized to assess the regressed parameters. The acquired errors of density calculations, in terms of the average absolute relative deviation percent (AARD%), were 1.210, 0.910, and 1.113% in the train, test, and total datasets, respectively, which confirms the excellent performance of the correlation. The developed general correlation was then compared to the ones from the literature that benefit from component-specific parameters. ANNs are also capable of calculating densities precisely. The best results of ANNs modeling were an error of 0.002% when the radial basis function (RBF) network was employed with the maximum number of neurons. It is inferred that the proposed correlation can work as a global equation to precisely estimate the densities of different AASs. In addition, ANNs offer a great performance to calculate the densities of AASs.