Prediction of excess enthalpy of polymer solutions by using ANN in comparison with extended NRTL and Wilson models

In this study, excess enthalpy of binary polymer solutions was represented by using the artificial neural network (ANN). Results were compared with extended NRTL and Wilson models. This research has used a multilayer feedforward network with Levenberg Marquardt backpropagation training for prediction of excess enthalpy. Original data were divided into two parts where 70% of data was used as training data and remaining 30% of data was used for testing. In this method, inputs of the artificial neural network are mole fraction and molecular weight. The number of neurons is set at six. Using ANN for prediction of excess molar enthalpies leads to less deviation compared to extended Wilson and NRTL equations and it is in good agreement with experimental data.

Keywords:

Prediction , Excess enthalpy , neural network , Wilson , NRTL , Polymer solutions