Simulation and determination of optimum conditions of pervaporative dehydration of isopropanol process using synthesized PVA–APTEOS/TEOS nanocomposite membranes by means of expert systems

Robust artificial neural network (ANN) was developed based on experimental data to predict dehydration of isopropanol by means of novel PVA–APTEOS/TEOS nanocomposite membranes in pervaporation (PV) process. The input properties were water concentration, feed temperature and nanoparticles content, while pervaporation separation index (PSI) was output. The Bayesian Regularization (BR) training method with full sampling was employed to train the network. Then, optimal ANN architecture was determined as 3:3:3:1 with log-sigmoid transfer function for hidden and output layers. The model finding revealed that nanoparticles content has significant effect on membrane performance (about 70%). The results demonstrated that the ANN model prediction and experimental data are quite match and the model can be employed with confidence for prediction of each nanocomposite membrane performance. Simulated annealing (SA) was also employed to determine controllable conditions to find the biggest PSI.

► Novel PVA–APTEOS/TEOS nanocomposite membranes were synthesized.
► ANN (in pervaporation) is rare until 2011 and the present work seems to be first.
► A review of ANN modeling used in different membrane process was included.
► The relative importance of operational conditions was determined.
► Optimum operational conditions and ANN architecture were determined.