Application of selected chemometric methods to describe and predict the properties of grafted ceramic membranes

•For designing new ceramic membranes, chemometric methods were used.•HCA and PCA were applied for modifiers similarity prediction.•The most important physicochemical properties of modifiers were evaluated.•Usefulness of chemometric methods for experimental results was also presented. Modification of a ceramic membrane surface for the purpose of hydrophobization can be achieved by the grafting technique involving the application of silane compounds and their derivatives. For the selection of the appropriate modifying compound, chosen chemometric methods can be implemented. In this study, the chemometric methods, i.e., hierarchical cluster analysis (HCA) and principal component analysis (PCA), were used for the evaluation of the properties of selected silane compounds applied as modifiers of ceramic membranes. The results of the calculations indicate that these methods are effective in predicting the properties of modifiers. Moreover, the PCA results provide additional information about the physicochemical properties of modifiers which influence grafted membrane properties. The data are essential when looking for (or synthesizing) new modifiers/linkers necessary for the preparation of ceramic membranes with the desirable properties It was found that physicochemical parameters such as molecular shape, atomic mass, total absolute charge, number of heavy atoms, polarizability, van der Waals volume, electronegativity as well as the number of H-bond acceptors are the most important. These parameters describe the observed properties of the grafting molecules and, simultaneously, the properties of the obtained modified ceramic membranes. It was also demonstrated that HCA and PCA can be used to interpret the previously obtained experimental results. For instance, HCA can limit the number of consecutive experiments aimed at estimating the stability of membranes in various organic solvents. Moreover, the applicability of multiple linear regression (MLR) analysis to predict the values of contact angle (CA) and water fluxes of modified membranes was also examined.