Genetic neuro-computing model for insights on membrane performance in oily wastewater treatment: An integrated experimental approach

Abstract

In this study, response surface methodology (RSM) and artificial neural network-based genetic algorithm (ANN-GA) were utilized to predict two crucial output parameters of membrane performance, namely separation efficiency and oil flux, derived from experimental investigations. The central composite design (CCD) screening approach of RSM was employed to evaluate the influence of important process input parameters, such as oil concentration (ranging from 50 to 10,000 ppm), feed flow rate (ranging from 150 to 300 mL/min), and pH of the feed (ranging from 4 to 10), as well as their synergistic effects on the output variables. The constructed RSM model and ANN-GA were effectively employed to estimate the optimum conditions for maximizing the output variables. Statistical analysis using the determination coefficient (R2) and standard error of prediction (SEP), along with analysis of variance (ANOVA) and the t-test, demonstrated the accurate description of the membrane performance process by both models. For the oil flux, the RSM model showed an estimated R2 of 0.9916 and SEP of 3.54%, while the ANN model exhibited an R2 of 0.9933 and SEP of 3.31%. In terms of separation efficiency, the RSM model yielded R2 = 0.9929 and SEP = 1.31%, whereas the ANN model achieved R2 = 0.9961 and SEP = 0.99%. Remarkably, the ANN-GA approach revealed the best optimum conditions for both responses. Furthermore, the sensitivity analysis of the developed ANN model indicated the order of significance of the variables as follows: oil concentration > feed pH > feed flow rate. These findings substantiate the efficacy of the proposed approach, making it viable for implementation in diverse industries to facilitate sustainable monitoring and management practices.