Performance analysis and optimization of sweep gas membrane distillation module using high-resolution computational model and machine learning

Abstract

Performance analysis of sweep-gas membrane distillation (SGMD) modules remains a challenging problem owing to the complex interaction and strong coupling between the flow parameters on the feed and permeate sides as well as the multiscale nature of gas permeation in the membrane pores. In this study, a framework for a practical optimization tool for SGMD was developed in which experimental data, numerical simulations, and machine learning (ML) techniques were systematically utilized. A high-resolution numerical simulation of the SGMD was introduced and experimentally validated for SGMD modules. The high-resolution numerical model was used to train a ML model with noticeable speedup and lower computational cost. The trained ML model was coupled with a multi-objective genetic algorithm to optimize the SGMD performance for a large-scale SGMD module with a novel and complex flow pattern. The SGMD, comprising five different modules with parallel sweep-gas patterns, showed interesting behavior in a sensitivity study based on modified Latin hypercube sampling. The sweep-gas temperature had the highest impact on the water permeation flux, which was due to the reheating of the feed flow as a result of the special configuration of the SGMD. Furthermore, the optimization solver successfully obtained the optimal design conditions for two different design scenarios by achieving a design target defined for an ammonia recovery ratio of 90% and an ammonia concentration of 25% at the sweep-gas outlet. The potential of the developed framework for practical engineering applications is highlighted.