(56d) Generalized Non-Linear QSPR Model for Liquid Viscosity

Liquid viscosity (LV) is an important thermophysical property involved in many chemical processes and phenomena. Although accurate models have been developed for classes of chemicals, a need for a generalized model exists. Further, empirically developed models require experimental data, which are costly and laborious to obtain. Therefore, there exists a need for generalized models capable of providing a priori predictions for compounds that have no relevant experimental data or for compounds that are yet to be synthesized.

Previously, researchers at Oklahoma State University (OSU) have developed a unified framework for correlating saturation properties. This scaled-variable-reduced-coordinates (SVRC) model is based on the corresponding states theory (CST) and scaling-law behavior and in general is capable of representing saturation properties within their experimental uncertainties. The modeling approach involves the use of quantitative structure-property relationship (QSPR) methodology to generalize the model parameters of the developed SVRC model. Specifically, we used SVRC to develop the behavior model, and QSPR to generalize the SVRC model parameters. This approach, in the past, has proven to be more effective than the typical efforts to develop generalized models directly using QSPR techniques. In this work, we extended the SVRC model to correlate LV values and generalized the model parameters using structure-property modeling. A database of 5,800 data points involving 335 fluids was used in the development of this model. The predictions on the LV values were within 4% of the modeling set values. When applied to an external dataset containing 16,450 data points involving 1235 fluids, the LV model predictions were, on average, within 9% of the reported values.