Application of QSPR to Binary Polymer/Solvent Mixtures: Prediction of Flory‐Huggins Parameters

Abstract A QSPR study was performed for the prediction of the Flory‐Huggins parameters of binary polymer/solvent mixtures. 1 664 descriptors for each polymer and solvent were checked and a cubic multivariable model, with R 2 = 0.9638 and s = 0.146, was produced by using genetic algorithms on a training set of 52 mixtures. The reliability of the proposed model was further validated by satisfactory statistical parameters being obtained using an external test set ( $R_{{\rm ext}}^2$ = 0.9565). All descriptors involved in the model can be derived solely from the chemical structures of the polymers and the solvents, which makes it very useful in predicting the Flory‐Huggins parameters of unknown or unavailable polymer/solvent mixtures. magnified image

Перевод пока недоступен