Development of Group Contribution Plus Property Models for Organic Systems
Advancing the chemical engineering fundamentals
Thermodynamics (T2-1P)
Keywords: UNIFAC, Group Contribution, Molecular descriptors, Connectivity Indices.
Properties of chemicals play a very important role in the design of the chemical product and in the design of chemical processes that can manufacture them. In computer-aided model-based methods and tools for synthesis and design of the product-process, the applicability of the model and its reliability is governed by the property models used in the product-process model. In addition to reliability of the property models, another important need in product-process synthesis and design is that the models need to be predictive and fairly simple and easy to use. Group contribution (GC) based property models satisfy most of the above requirements and are therefore, routinely used in product-process synthesis and design. There are however some limitations, for example, they are not very reliable for complex chemicals or systems of complex chemicals; group parameters may not be available; not enough experimental data is available to develop a large set of group parameters; reliability is often dependent on how the chemical structure is represented by the groups. These limitations make the synthesis and design of high-value chemical products such as drugs, pesticides, structured polymers, etc., difficult with model-based methods since the necessary reliable property models are not available.
A complementary knowledge based on atomistic models, molecular descriptors and other techniques can be helpful to overcome the inherent limitations of the conventional group contribution methods. For example, features of the molecules like steric effects due to spatial conformation can be safely correlated using hybrid models that combine the simplicity and the predictive nature of GC-models with basic molecular/atomic knowledge for the property and model under study. This hybrid models (called GC-Plus models) should be simple enough to predict properties in an automatic fashion, and accurate in order to improve the performance of the conventional GC models.
This paper presents the current state of hybrid models for property prediction in organic systems: GC-Plus models have been developed for the automatic generation of missing group contributions (for pure compound property prediction) and for the generation of missing group interaction parameters for mixture property prediction models (UNIFAC-CI). Emphasis is given to the increase of the applicability range, specifically solid-liquid equilibria (SLE) calculations for mixtures (UNIFAC-CI) and accurate models for pure component property prediction of stereoisomers. The development of this hybrid models is based on molecular descriptors such as connectivity indices (CI) and molecular simulation methods (force fields).
Choosing properly the molecular/atomic based complement for the Group Contribution models -in terms of computational time- will derive in automated property models able to generate new and useful data for process simulation. Besides this, will be possible to fine-tune the parameters related with the models and to modify this models according the needs of the user. This property model development is the basis of a systematic computer aided framework that guides the user to create/select the appropriate property model, test its validity and then use it for the necessary calculations.
Case studies for problems from industry will be shown in order to exhibit the improvement gained in property prediction using GC-Plus models for organic systems involving either pure compounds and mixtures. The prediction of phase equilibria involving mixtures uncovered for the UNIFAC group contribution method and pure compound property prediction that are important in high-value chemical product design will be highlighted using the computer aided framework.
Presented Monday 17, 13:30 to 15:00, in session Thermodynamics (T2-1P).
