| Modeling the phase behavior of substances using thermodynamic equations of state (EOS) has been a vital area of research, as it is a quick and inexpensive way of predicting the phase properties at desired conditions. The parameters obtained after parameterising the pure components are used along with appropriate mixing rules to predict the mixture phase behavior of these compounds. Prior optimization techniques used for parameterization of pure components are methods that converge at a local minimum. These parameters (local minimum) depend on the conditions of the optimization such as the objective function used, number and weighting of experimental data points, etc. In this work, a systematic analysis of the effect of these factors on pure component parameterizations as well as on mixture predictions was investigated. It was found that these factors have a large impact on the quality of the mixture phase predictions.; Thereby, this work attempts to characterize and maximize predictive power of the pure component parameters by mitigating spurious conclusions that are based on results from local minimization schemes and, thus, allow for more definitive conclusions on the properties of systems in the absence of experimental data. To this end, this work employs global terrain methodology---an advanced global optimization technique and applies it to the field of thermodynamic modeling. Additionally, analyses on the effects of multiple parameter-sets on the prediction of mixture properties will be discussed. Through this work, parameter rules will be developed that will allow for the optimal prediction of pure component properties. |
