Improving the Predictability of Chemical Equilibrium Software

Over the past few decades, researchers have been developing tools to predict chemical reactions to aid the growing field of industrial chemistry. Currently, a large variety of numerical tools are used to predict the final chemical equilibrium based on the minimization of the Gibbs free energy. Because of the mathematical complexity of the problem, numerical methods were developed to solve this problem. These methods were reviewed in another study (submitted for publication in Comput. Chem. Eng., Predicting multi-phase chemical equilibria using a Monte Carlo technique, 2018) exhibiting their limitations and proposing an alternative. In this study, the sensitivity of the prediction as a function of the thermochemical (input) parameters is discussed showing that significant deviations are possible when the relative uncertainty between the enthalpies of formation is larger than a few kJ/mol. Often the scatter between various data sources is much larger than this. To solve this difficulty, it was attempted to derive all the required thermodynamical parameters from a base of molecular descriptors common to the chemistry targeted in this work (organic). The group contribution theory is implemented and in particular the UNIFAC descriptors and is shown to give very satisfactory results