AI-PCSAFT approach: New high predictive method for estimating PC-SAFT pure component properties and phase equilibria parameters

Abstract

In this work, a new approach based on the association of Artificial intelligence method (AI) and PC-SAFT equation of state is applied to conceive a model for estimating the solubility of solid drugs in supercritical carbon dioxide. Neuro-equation of state approach (NES) is the new technique that takes benefit from the advantages of both ANN and PC-SAFT equation of state. The new method decomposes into three main stages, first the optimization of direct ANN for predicting solids-scCO2 phase equilibria (where 15 binary systems are used), then the ANN inverse is performed to be an alternative to group contribution methods (GCMs) for estimating the pure components and physical properties (reduce the uncertainty committed in estimating these properties) and enhance the PCSAFT equation of state to estimate phase equilibria parameters and finally, ANN-PCSAFT approach is used to estimate the solubility of 213 solid solutes in supercritical carbon dioxide. The performance strategy has been carried out using a linear regression analysis of the predicted versus experimental outputs, as an indication of the predictive ability of the developed method.The new approach is successfully applied to the phase equilibria modeling for 213 binary systems with high accuracy (the comparison in terms of average absolute relative deviation (AARD %) showed a variation from 2 to 6%) and allowed to enhance the phase equilibria modeling by reducing the number of optimized parameters and surpass the main drawbacks faced in this area mainly the non-availability of physical properties and EOS pure component properties and the limitation of the equation of state.