Abstract
In this work, a novel hybrid model was developed in order to study the membrane-based liquid separation process. The membrane system was a continuous hollow-fiber membrane module for contacting two aqueous and organic phases for reactive extraction of benzoic acid (BA) from aqueous solution. Two simulation approaches were utilized in order to build a robust hybrid model. The hybrid model is composed of computational fluid dynamics (CFD) and Adaptive Neuro-Fuzzy Inference System (ANFIS) elements. First, the CFD approach was used in order to capture the mass transfer of the system, whereas ANFIS was trained using the obtained CFD results. The hybrid model was used to predict the concentration distribution of solute in the membrane contactor. The combined simulation methodology can reduce the computational costs and time significantly, while it predicts the process with high accuracy. The ANFIS was trained based on the extracted data of concentration distribution from the CFD simulations, and the training and test analyses indicated great agreement. Different membership functions were evaluated, and it was revealed that using three functions, an {R^{2}} of 0.996 was obtained. The simulation results reveal that the BA concentration was changed along the membrane length and diffusional mass transfer is more significant in order to improve the separation efficiency of BA using membrane contactors. The developed hybrid simulation methodology is capable of design and optimization of membrane-based separation at low computational expenses and provides a predictive tool for process intensification.