Evaluation of surface rheology parameters of two nonionic surfactants by implementation of Machine Learning Techniques: a new approach of study

Computational approaches are constantly being developed and implemented in fields such as chemical and engineering, in order to better understand their different standards and efficiently bringing them a quick computational solution, in which it avoids exhaustive and countless experimental assessments, based on similar systems studied before. However, regarding interfacial phenomena, studies on foam behavior and surface rheology are limited, although both fields have substantial body of research and published papers. Hence in the present work, a new approaching tool was implemented to study the behavior of two nonionic surfactants: one sugar-based and one glycol based. It consisted of systematic applications of different regression Machine Learning techniques, named Decision Trees (DT), Random Forest (RF) and XGBoost (XGB), on a dataset collected from different works published on foam and surface rheology. The performance of these models was evaluated by two different metrics (RMSE and R²). The outcomes showed a good performance, with remarkably low error rates for all the three models concerning surface rheology experiments, with the lowest R² reaching 0.970 for DT model, while XGB model provided the best-performing prediction, with R² reaching 0.985. Furthermore, Pearson’s correlation coefficients and feature importance scores were also used to further assess the significance of the constants related to the studied properties. From an industrial point of view, we expect that we will start a new path in colloidal scientific research in which thousands of experiments will not be needed to predict a new system behavior.