Modeling Study on the Density and Viscosity of Ionic Liquid–Organic Solvent–Water Ternary Mixtures

The accurate prediction of physical properties is critical
for
the successful application of both conventional and novel chemicals
across various industries. This work focuses on predictive modeling
for the density and viscosity of ternary mixtures of ionic liquids
(ILs) using a combination of the group contribution (GC) method and
three machine learning algorithms: artificial neural network (ANN),
XGBoost, and LightGBM. Initially, a comprehensive collection of reliable
open-source data is compiled, comprising 10,553 data points from densities
for 28 classes of ILs and 33 classes of organic solvents (os) and
3581 data points from viscosity for 15 classes of ILs and 17 classes
of os. The modeling results demonstrate that all three machine learning
algorithms yield reliable predictions. Notably, the ANN-based model
showed the best performance in both density and viscosity property
predictions, with a density fit of more than 0.99 and a viscosity
fit of more than 0.98. To gain a deeper understanding of the influencing
factors, the study employed the Shapley Additive Interpretation (SHAP)
technique. This study provides valuable insights into accurately predicting
two important properties of IL–organic solvent–water
ternary mixtures. By enabling more efficient screening of IL–os–water
mixed solvents in industrial design, these findings contribute to
the advancement and optimization of IL-based processes across various
applications.