Artificial neural network modeling and response surface methodology of desalination by reverse osmosis

Response surface methodology (RSM) and artificial neural network (ANN) have been used to develop predictive models for simulation and optimization of reverse osmosis (RO) desalination process. Sodium chloride aqueous solutions were employed as model solutions for a RO pilot plant applying polyamide thin film composite membrane, in spiral wound configuration. The input variables were sodium chloride concentration in feed solution, C. feed temperature, T, feed flow-rate. Q, and operating hydrostatic pressure, P. The RO performance index, which is defined as the salt rejection factor times the permeate flux, has been considered as response. Both RSM and ANN models have been developed based on experimental designs. Two empirical polynomial RSM models valid for different ranges of feed salt concentrations were performed. In contrast, the developed ANN model was valid over the whole range of feed salt concentration demonstrating its ability to overcome the limitation of the quadratic polynomial model obtained by RSM and to solve non-linear problems. Analysis of variance (ANOVA) has been employed to test the significance of response surface polynomials and ANN model. To test the significance of ANN model, the estimation of the degree of freedom due to residuals has been detailed. Finally, both modeling methodologies RSM and ANN were compared in terms of predictive abilities by plotting the generalization graphs. The optimum operating conditions were determined by Monte Carlo simulations considering: (i) the four input variables, (ii) for typical brackish water with a fixed concentration of 6 g/L and (iii) for typical seawater with a fixed concentration of 30 g/L. Under the obtained optimal conditions maximum RO performance indexes have been achieved experimentally. (C) 2010 Elsevier B.V. All rights reserved.