Artificial Intelligence based modeling of pervaporation process for alcohol dehydration

Abstract In this study, an Artificial Intelligence/artificial neural network (ANN) based approach is used to model the pervaporation process for alcohol dehydration using an inorganic membrane. The Deep Learning Toolbox of MATLAB was used to model an artificial neural network. ANN was developed with a single hidden layer and a variable number of neurons. Results showed that minimum Mean Squared Error was found with 12 neurons for permeate flux and 15 neurons for selectivity, and thus, the same number of neurons were considered for further simulations. Model predictions showed a good fit with experimental results with 0.96 and 0.98 correlation factors for permeate flux and selectivity, respectively, which provide confidence in the model predictions. Later, trained neural networks were used to study the effects of operating parameters. A maximum value of flux 5.8 kg/m2h was found at 90 °C feed temperature and 2.9 kPa permeate pressure. Similarly, the highest selectivity for water 1099 was found at 100 dm3/h flow rate and 0.96 kg/kg isopropanol concentration.