Artificial neural network modeling and optimization of desalination by air gap membrane distillation

Abstract

An experimental based artificial neural network (ANN) model is constructed to describe the performance of air gap membrane distillation process for different operating conditions. The air gap thickness, the condensation temperature, the feed inlet temperature, and the feed flow rate of salt aqueous solutions are the input variables of this process, whereas the response is the performance index, which takes into consideration both the permeate flux and the salt rejection factor. The neural network approach was found to be capable for modeling accurately this membrane distillation configuration. The overall agreement between the ANN predictions and experimental data was very good showing a correlation coefficient of 0.992. To test the statistical significance of the developed ANN model the analysis of variance (ANOVA) has been employed. According to ANOVA test, the ANN model is found to be statistically valid and can be used for the prediction of the performance index. Finally, the predictive abilities of the ANN model were ascertained by plotting the 3D generalization graphs. The optimum operating condition was determined by Monte Carlo stochastic method and the obtained optimal conditions are 3.0 mm air gap thickness, 13.9 degrees C condensation temperature, 71 degrees C feed inlet temperature and 205 L/h feed flow rate with a maximum experimental performance index of 51.075 kg/m(2) h and a residual error less than 1%.