- 2:00 PM

Increased Application Range of Property Models without New Experimental Data

Hugo E. González1, Jens Abildskov1, Rafiqul Gani2, Tariq A. Khan3, Sabyasachi Sen4, and Bluck David4. (1) Chemical and Biochemical Engineering, Technical University of Denmark, Sølftofts Plads bygn. 227, Kgs. Lyngby, Denmark, (2) CAPEC, Department of Chemical and Biochemical Engineering, Technical University of Denmark, Søltofts Plads, Building 229, 2800 Kgs. Lyngby, Denmark, (3) SimSci-Esscor - Global Solutions Development, Invensys Development Center India Pvt. Ltd., Plot #17, Orion Building, BLock - J, VIT Park, Software Units Layout - Madhapur, Hyderabad, India, (4) Invensys SimSci-Esscor, 26561 Rancho Parkway South, Suite 100, Lake Forest, CA 92630

Properties of chemicals are fundamental for the design and analysis of chemical, pharmaceutical, food, agrochemical and related industries. In order to meet the increased demands with respect to complexity of the chemical molecular structures, wider range of chemicals and accuracy, further development of current estimation methods, techniques and/or development of new models are necessary together with additional experimental data. The objective of this paper is to present a new combined group contribution-atom connectivity modelling approach that is able to extend the predictive application range of property models (pure component as well as mixtures). The paper will provide details of the development of a hybrid model that combines molecular descriptors theory and group contribution theory for pure component properties and for mixture properties prediction using a GCPlus approach[1] . In this paper the Original UNIFAC and UNIFAC-Dortmund based mixture property prediction for phase equilibria calculations will be highlighted together with new developments in the pure component properties and polymer repeat unit properties.

The main idea of this methodology is the use of connectivity indices (CI) to describe the molecular fragmentation that is characteristic for the UNIFAC group contribution method and that relates properties (molecular interactions in this case) with molecular structure. The result is the automatic generation of group interaction parameters (GIPs) for the UNIFAC group contribution methods. Derivation and use of the method has been reported in a recent publication[2]. There, the main features of the GCplus approach used on UNIFAC are discussed: comparisons of the performance using the generated GIPs versus original UNIFAC GIPs (when the GIPs are available) vs experimental data (when the GIPs are not available) and the performance of original UNIFAC using group interactions including a totally new generated group to describe a molecule. So far, the current status of the GCPlus approach for mixtures includes 1) Original UNIFAC: groups formed by C, H, O, N, Cl, and S atoms for VLE; extrapolation of the VLE GIPs to SLE calculations; a set of GIPs for LLE using groups containing C,O,H atoms and 2) UNIFAC-Dortmund: GIPs for groups containing C,O,H, atoms. This feature on the methodology implies the potential generation of a very large number of GIPs as well as filling-up the GIPs matrices without the need for additional experimental data. In the pure component properties area, models for viscosity, surface tension, Hansen solubility, parameters have also been recently developed. The GCPlus approach is simple and easy to implement/use.

It will be highlighted, the use and performance of original UNIFAC and UNIFAC-Dortmund with GIPs generated through CI for mixtures including heteroatoms and different classes of phase equilibria, an analysis of the correlation errors and the evaluation of prediction ranges. Performance of the predicted GIPs en SLE and LLE calculations will also be highlighted. Finally, the scope and significance of the GCPlus approach will be highlighted for cases where no parameters and/or experimental data exist, thereby providing a true and reliable predictive power to properties estimation in process-product design and synthesis.


1 H. E. González, J. Abildskov, R. Gani (2007) Computer-aided Framework for Pure Component Properties and Phase Equilibria Prediction for Organic Systems, Fluid Phase Equilibria, 261, (1-2), 199-204.

2 H. E. González, J. Abildskov, R. Gani, P. Rosseaux, B. Le Bert (2007) A Method for Prediction of UNIFAC Group Interaction Parameters, AIChE J., 53:1620-1632.