Modeling and optimization of membrane fabrication using artificial neural network and genetic algorithm

In the present study, the process of membrane preparation was successfully modeled by artificial neural network (ANN). In the experimental section, polyethersulfone (PES) and polysulfone (PS) membranes were prepared by immersion participation using different types of solvent including N,N-dimethylacetamide (DMAc), non-solvent such as 2-propanol (IPA) and additives including polyvinylpyrrolidone (PVP). The mechanical, chemical and thermal properties of polymers, solvent, non-solvent and additives were elucidated for introducing the ANN. The effects of polymer and additive concentrations on membrane performance were investigated by validated ANN. The optimum concentrations to achieve maximum flux for quaternary systems (PES/PVP/DMAc/IPA + water) and (PS/PVP/DMAc/IPA + water) were estimated by genetic algorithm and assembled ANN. For PES membrane, a good agreement was found with experimental data and SEM micrographs with relative error of 5% and 1% for flux and rejection.

Research highlights▶ Modeling of membrane preparation is possible using artificial neural network. ▶ Genetic algorithm is able to estimate optimum preparation conditions using ANN developed model. ▶ Good agreement was found between model and experimental data with 5% error.