QSPR prediction of surface tension of refrigerants from their molecular structures

In this work, quantitative structure-property relationship (QSPR) models for prediction of surface tension of 224 refrigerant compounds on the basis of their molecular structures were developed by using genetic function approximation (GFA) and adaptive neuro-fuzzy inference system (ANFIS) methods. GFA was used to select the most important molecular descriptors and develop the linear model. To develop a nonlinear model, the four descriptors selected by GFA were used as the inputs for ANFIS method. The predictive ability of the developed model was evaluated by predicting the surface tension of a number of compounds as a test set. The squared correlation coefficients of surface tension predicted by the GFA and ANFIS methods were 0.985 and 0.996, respectively. The final results suggest that the obtained QSPR model can be applied for predicting the surface tension of refrigerant compounds with high accuracy and simplicity.

► A reliable QSPR model was developed for estimation surface tension of refrigerants. ► The developed QSPR model by GFA and ANFIS methods contains only four descriptors. ► This QSPR model can easily be applied with high accuracy (R2 > 0.985). ► The major affecting factors on the surface tension of refrigerants were considered.